segment_id
stringlengths
13
56
⌀
video_id
stringlengths
11
92
title
stringlengths
8
102
company
stringclasses
5 values
channel
stringclasses
12 values
video_type
stringclasses
2 values
upload_date
stringclasses
834 values
duration_seconds
float64
1.21k
18k
⌀
url
float64
description
stringlengths
0
4.76k
topics
listlengths
2
5
qc_flags
float64
segment_index
int64
0
85
⌀
total_segments
int64
1
86
⌀
heading
stringclasses
997 values
heading_confidence
stringclasses
2 values
start_time
stringlengths
4
7
⌀
text
stringlengths
2
73.9k
⌀
word_count
int64
1
14k
⌀
fAf4Zh-CC08_12
fAf4Zh-CC08
Boost AI context with hybrid search in Spanner
Google
Google Cloud Tech
public
2026-06-25
null
null
Go beyond keyword search to get the most out of your operational data. In this session, learn to build sophisticated, AI-powered experiences with Spanner. Discover how to combine full-text and vector search into powerful multi-model queries that deeply understand user intent. We’ll unveil new AI innovations that simpli...
[ "AI / Machine Learning", "Gemini / Google AI", "RAG & Vector Search", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
12
21
null
null
0:00
hair dye in Google, we do this exact same thing. And what search allows you is it brings this power to your product search. All right. The second component of AI Search is vector search. So while in full text search you're basically tokenizing your documents and you're in the realm of words. What vector search allows y...
325
fAf4Zh-CC08_13
fAf4Zh-CC08
Boost AI context with hybrid search in Spanner
Google
Google Cloud Tech
public
2026-06-25
null
null
Go beyond keyword search to get the most out of your operational data. In this session, learn to build sophisticated, AI-powered experiences with Spanner. Discover how to combine full-text and vector search into powerful multi-model queries that deeply understand user intent. We’ll unveil new AI innovations that simpli...
[ "AI / Machine Learning", "Gemini / Google AI", "RAG & Vector Search", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
13
21
null
null
0:00
this even to generate your vector embeddings. So basically you could have an ML model sitting in Vertex or Gemini enterprise. And then you could basically have Spanner directly call your model, generate the embeddings and keep them up to date and build an index. It all works out of the box with DDL. We also have integr...
325
fAf4Zh-CC08_14
fAf4Zh-CC08
Boost AI context with hybrid search in Spanner
Google
Google Cloud Tech
public
2026-06-25
null
null
Go beyond keyword search to get the most out of your operational data. In this session, learn to build sophisticated, AI-powered experiences with Spanner. Discover how to combine full-text and vector search into powerful multi-model queries that deeply understand user intent. We’ll unveil new AI innovations that simpli...
[ "AI / Machine Learning", "Gemini / Google AI", "RAG & Vector Search", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
14
21
null
null
0:00
are combining your vector search with some other are filters. So for instance, when Gmail searches all the mails for an individual user, that's a small result set right. And so in those cases KNN works great. It works well for small corpuses highly filtered result sets and partition queries. But when you want to scale ...
325
fAf4Zh-CC08_15
fAf4Zh-CC08
Boost AI context with hybrid search in Spanner
Google
Google Cloud Tech
public
2026-06-25
null
null
Go beyond keyword search to get the most out of your operational data. In this session, learn to build sophisticated, AI-powered experiences with Spanner. Discover how to combine full-text and vector search into powerful multi-model queries that deeply understand user intent. We’ll unveil new AI innovations that simpli...
[ "AI / Machine Learning", "Gemini / Google AI", "RAG & Vector Search", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
15
21
null
null
0:00
it will use the Euclidean distance in this case to traverse the index, find the most relevant clusters, and do a localized search. So essentially, what you can see here is you're trading off a little bit of precision by say not looking at all the documents but just a subset. But then this becomes much faster and much c...
325
fAf4Zh-CC08_16
fAf4Zh-CC08
Boost AI context with hybrid search in Spanner
Google
Google Cloud Tech
public
2026-06-25
null
null
Go beyond keyword search to get the most out of your operational data. In this session, learn to build sophisticated, AI-powered experiences with Spanner. Discover how to combine full-text and vector search into powerful multi-model queries that deeply understand user intent. We’ll unveil new AI innovations that simpli...
[ "AI / Machine Learning", "Gemini / Google AI", "RAG & Vector Search", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
16
21
null
null
0:00
a table called products. And you have this query term space themed logo for a+ So the full text search is basically looking at all your items going through the index and finding documents that match. So it's basic token based matching right. So you get your text matches in the middle. And then instead of using a docume...
325
fAf4Zh-CC08_17
fAf4Zh-CC08
Boost AI context with hybrid search in Spanner
Google
Google Cloud Tech
public
2026-06-25
null
null
Go beyond keyword search to get the most out of your operational data. In this session, learn to build sophisticated, AI-powered experiences with Spanner. Discover how to combine full-text and vector search into powerful multi-model queries that deeply understand user intent. We’ll unveil new AI innovations that simpli...
[ "AI / Machine Learning", "Gemini / Google AI", "RAG & Vector Search", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
17
21
null
null
0:00
do they have multiple retrievers and some reranking logic to figure out how to merge the results. Finally, graph. I hope many of you had a chance to go listen to our talk earlier in the day. If you haven't, I would urge you to go check it out later. So Spanner has this really nice implementation of graph. So the way we...
325
fAf4Zh-CC08_18
fAf4Zh-CC08
Boost AI context with hybrid search in Spanner
Google
Google Cloud Tech
public
2026-06-25
null
null
Go beyond keyword search to get the most out of your operational data. In this session, learn to build sophisticated, AI-powered experiences with Spanner. Discover how to combine full-text and vector search into powerful multi-model queries that deeply understand user intent. We’ll unveil new AI innovations that simpli...
[ "AI / Machine Learning", "Gemini / Google AI", "RAG & Vector Search", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
18
21
null
null
0:00
of approach is it's relying on the model's understanding of the relationships in your data. So the ML model that generates your embedding needs to understand the relationships in your data. And so you're going through that bottleneck the ML model bottleneck to model what you already know about your data and try to teac...
325
fAf4Zh-CC08_19
fAf4Zh-CC08
Boost AI context with hybrid search in Spanner
Google
Google Cloud Tech
public
2026-06-25
null
null
Go beyond keyword search to get the most out of your operational data. In this session, learn to build sophisticated, AI-powered experiences with Spanner. Discover how to combine full-text and vector search into powerful multi-model queries that deeply understand user intent. We’ll unveil new AI innovations that simpli...
[ "AI / Machine Learning", "Gemini / Google AI", "RAG & Vector Search", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
19
21
null
null
0:00
are some other talks that we have on Spanner. We basically I think most of these have completed. But there is a very interesting one tomorrow if you're interested where we are going to talk about some new analytics capabilities that we've just added to Spanner. It's a very cool talk. I would really urge you to check it...
325
fAf4Zh-CC08_20
fAf4Zh-CC08
Boost AI context with hybrid search in Spanner
Google
Google Cloud Tech
public
2026-06-25
null
null
Go beyond keyword search to get the most out of your operational data. In this session, learn to build sophisticated, AI-powered experiences with Spanner. Discover how to combine full-text and vector search into powerful multi-model queries that deeply understand user intent. We’ll unveil new AI innovations that simpli...
[ "AI / Machine Learning", "Gemini / Google AI", "RAG & Vector Search", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
20
21
null
null
0:00
Spanner also index using an NSWT kind of methodology or is it purely. So right now we mainly have scan which is the clustering mechanism I told you about. I think we have some plans to take up NSWT, but we should talk offline. Yes this is something that we are exploring as well. Thank you. Yeah thank you. All right. Th...
104
05043f3GseE_0
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
0
23
null
null
0:00
Good afternoon, everyone. If you are here because you're interested in how Google DeepMind is doing innovation and how Google Cloud is bringing that to you are in the right place. My name is. I am the product manager for all the model capabilities that are available on the Gemini Enterprise Agent platform. Today, I'm g...
325
05043f3GseE_1
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
1
23
null
null
0:00
for all of you. So to really understand the platform, for those of you who didn't see the keynote, I think it's important to understand how software engineering is developing over time. In the past, this is my personal opinion. Machine learning was always actually about adding agency into software. It was just at a dif...
325
05043f3GseE_2
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
2
23
null
null
0:00
make things better for you as our users, what else can we do to make things even more optimal. Well, one solution might be why don't we just create the agents on your behalf instead of you thinking about building an agent and scaling it and governing it. Can we do that. Can Google do that. Because we built the model, w...
325
05043f3GseE_3
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
3
23
null
null
0:00
to think about using any sort of Gemini based agent that Google is creating for you. Well, one of the reasons is we know how to really work with this model and get the best out of it. This new API is not only new functionality in an API interface, it is also improvements in the quality of the way we perform Deep Resear...
325
05043f3GseE_4
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
4
23
null
null
0:00
our Google DeepMind team to get hands on, see the details as well as do some demos. Hi, everyone. My name is Philip. I'm part of the Google DeepMind team working on a Gemini API and also on the Deep Research API. So I will start a little bit talking about how we build the agent, what it does, how it works, and then we ...
325
05043f3GseE_5
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
5
23
null
null
0:00
And after the research is done, we have a separate synthesized process which creates those nice little reports and also now with the visuals and the results speak for themselves. So the new Gemini 3 point, I'm sorry, the Gemini Deep Research preview agent or Max agent achieves 55.6 on humanities last exam. That's basic...
325
05043f3GseE_6
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
6
23
null
null
0:00
agriculture, there are now specific providers which allows you to connect your data to those servers. And then also the agent is built with resiliency in mind, like Deep Research requests run not just a few seconds. They oftentimes take minutes or even hours to complete. So we solved this by managing the state on a ser...
325
05043f3GseE_7
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
7
23
null
null
0:00
Is it already done. Or you can also enable streaming to get thought summary updates while the request is running. So we already quickly talked about the capabilities, but here's a quick overview on what was available until yesterday basically, and what is available today or then starting in a few days. So it's availabl...
325
05043f3GseE_8
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
8
23
null
null
0:00
over the plan before starting the research. And it has support for remote MCP servers and we are not working alone. Basically, we have many great partners, especially from the financial ecosystem. Factset is also here. Patrick will speak after me to show you a little bit more on how they used Gemini Deep Research, how ...
325
05043f3GseE_9
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
9
23
null
null
0:00
Google Search and URL context. And we will start a query. And then we will see if it completes until the demo is done. If not, I have something prepared. So we can start. Very simple. Maybe we can research the topic the new features from Google Cloud Next 2026\. Let's see how good our research will be. And now what hap...
325
05043f3GseE_10
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
10
23
null
null
0:00
you started, it's also available if you prefer JavaScript or another language to use the Search API with it. In this case, I will use Python. We will get our Gemini API key. We install the latest Google Chennai SDK. That's our Python SDK to get started. That's also the same SDK which you will use with the Gemini Enterp...
325
05043f3GseE_11
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
11
23
null
null
0:00
same request. Here we use the previous interaction ID. This basically means we append to the previous request and we say, hey, still collaborative planning to add please a section on a total cost of ownership. And then we wait a bit and assuming our plan is OK, that's something we want to go with. We turn the collabora...
325
05043f3GseE_12
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
12
23
null
null
0:00
It also supports real time streaming. We have seen it. Maybe we can quickly look back to our OK, that's cool. Our initial request was completed on the new features from Google Cloud. And now the report here took five minutes. So it really depends based on how deaf your query is, how long the agent will work, the overvi...
325
05043f3GseE_13
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
13
23
null
null
0:00
own rug store with inside of Gemini. And for Gemini agent enterprise platform, you can connect it to your Vertex AI store. So file search is on the Gemini API. That's same as the files I search on Vertex previously, and then also for providing images and/or multi-modal data. You can provide your text prompt, but you ca...
325
05043f3GseE_14
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
14
23
null
null
0:00
on how those visuals could look. So here we don't have a chart. We really have something more concrete, a pie chart and other visuals, which we will see later a bit more and we have a great documentation. So if you want to read about it more, if you want to dive deeper, if you don't use Python, if you use JavaScript or...
325
05043f3GseE_15
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
15
23
null
null
0:00
leaders in multi-asset class financial data. We started as a content aggregator 45 years ago. We didn't collect our own data, but we built great user experiences on top of that information for our users. And that has really allowed us to become a trusted platform for investment professionals around the world. And as we...
325
05043f3GseE_16
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
16
23
null
null
0:00
So how do we go about doing that. So the most important part is how do we source all of the content. How does the trusted data come together. This is one of our secret sources, and it comes from how we ingest huge volumes of documents and process them and turn them into the data products that flow into our solutions ev...
325
05043f3GseE_17
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
17
23
null
null
0:00
solutions that our clients need. And then we align all that content across a whole other data ecosystem that comes from exchanges and other third parties that participate in the financial space. And that data all needs to be aligned. And historically we've delivered across software meant for human users, things like ou...
325
05043f3GseE_18
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
18
23
null
null
0:00
the same type of capabilities. So let's think of the workflow that a lot of investment professionals have. I'm riding on the train in the morning. I'm going to the office. My day is full of client meetings. When I'm listening to a podcast, maybe that I've generated off of another great research tool, and I hear a new i...
325
05043f3GseE_19
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
19
23
null
null
0:00
me. Tell me if you think I'm right. If you think I'm wrong and where else I can go and research this topic a little more deeply. Fantastic type of workflows. And so as these research reports go and run, they can build these visuals. They can help me understand the walls and walls of data that are going to come out of t...
325
05043f3GseE_20
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
20
23
null
null
0:00
data. Let's scrape the web, let's look at our other internal sources and let's see where does it line up, where does it not. The dead end type of handling and the synthesis across time frames and time horizons. Different data types are really powerful. So that's where the next horizon we think with these models lies is...
325
05043f3GseE_21
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
21
23
null
null
0:00
that require evaluating all of the scientific literature, and it lets us answer questions in days rather than waiting weeks or months. What Deep Research is able to do in an impressive way is connect that quantitative data to more qualitative data, and really bring that to life. Taxa was an early adopter to a more open...
325
05043f3GseE_22
05043f3GseE
Implementing DeepMind innovation: Deep research API
Google
Google Cloud Tech
public
2026-06-25
null
null
In this session, learn how Google DeepMind develops agents powered by Gemini that help improve performance and complete comprehensive research tasks. And discover how you can use Google Cloud to implement these techniques in your applications. Speakers: Advait Bopardikar, Philipp Schmid, Patrick Starling
[ "Gemini / Google AI", "Agents & Autonomous Execution", "Cloud Architecture", "Developer Tools" ]
null
22
23
null
null
0:00
for so long. Deep Research has really accelerated our team. It saved us from having to build out really complex workflows. Let's our scientists iterate extremely quickly, and it lets them focus on what questions they want to ask. The number of ideas that could come to fruition, the quality of those ideas, and the way t...
90
yVvLjWDmb0Y_0
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
0
19
null
null
0:00
All righty. Hello, and welcome everyone to. Engineering the future of Kubernetes for AI at scale. Today we're going to be talking about how Google is working with the open source community to drive industry wide standardization and transformation. I'm Kaslin Fields and I run GKE developer relations engineering here at ...
325
yVvLjWDmb0Y_1
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
1
19
null
null
0:00
your business. This is great in a world where we have a lot of stateless applications that run on CPUs and don't have to complex of memory and state requirements where something needs to make sure that it has the same IP address all the time. For example, Kubernetes autohealing capabilities. Often traditionally have me...
325
yVvLjWDmb0Y_2
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
2
19
null
null
0:00
and fast. Sure, we introduced stateful sets 10 years ago, but the challenge of high performance data access and management in a distributed system continues. So today, we're going to highlight what we're doing in Kubernetes to address the challenges of today's AI workloads. Some of you may be sitting here thinking, is ...
325
yVvLjWDmb0Y_3
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
3
19
null
null
0:00
we'll talk about model servers and how you do that. The other is AI apps and agents. So these are programs, actual code that runs on computers. That takes a dependency on models. They're programs that use non-determinism that's offered by AI models to achieve some goal. And as far as we're concerned, in Kubernetes land...
325
yVvLjWDmb0Y_4
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
4
19
null
null
0:00
agents. And in the demo, in the keynote, the presenter, Frank hinkelmann, who is also in DevRel at Google Cloud, described agents this way. And I think about this statement a lot. I think it's a really useful framework for understanding what an agent is. She said all you need to build an agent is instructions, tools, a...
325
yVvLjWDmb0Y_5
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
5
19
null
null
0:00
the magical non-determinism math or a model. So a model is just math. Math that does things that you can't anticipate. It's non-deterministic. You give it a problem and it figures it out, possibly by realizing that it needs to use some tools. We'll talk more about models and how to run them later. But this is basically...
325
yVvLjWDmb0Y_6
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
6
19
null
null
0:00
who's seen the web UI. Who hates it. Sorry about that. I didn't make it, but still sorry about it. So this is the web UI. When you create an agent with Development Kit, it's a framework that basically you specify which model you're using, what the instruction is you want to give it and what tools it has access to. And ...
325
yVvLjWDmb0Y_7
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
7
19
null
null
0:00
object that the custom resource definition allows you to define in your cluster. You first define a sandbox claim template that says if I had a sandbox, I would want it to look like this. And then when you need one, it makes a sandbox claim that adheres to that template. So you can see here I've got a couple of them. T...
325
yVvLjWDmb0Y_8
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
8
19
null
null
0:00
up a whole bunch of copies of the same snapshotted really fast. The most common use case that I get when I talk with PMs or engineers who are working on this about it is models actually, which this is not in the model section, but when you need to spin up a model, you need to get the weights for that model in order to ...
325
yVvLjWDmb0Y_9
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
9
19
null
null
0:00
that you can make use of within place. Pod resize. I also want to mention here Horizontal Pod autoscaling is very well known, where you create more copies of a workload so that you have more to load balance against. Vertical, auto or autoscaling is where you resize the requests and limits on the. It used to be that if ...
325
yVvLjWDmb0Y_10
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
10
19
null
null
0:00
up and down, up and down very quickly. All it takes for this to work is to add the subresource equals resize if you're running it by kubectl. This is over here. And you can see I just randomly played with a 0 restarts just changing the requests. And this is for me it's pure magic. Consider millions of pods. Every small...
325
yVvLjWDmb0Y_11
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
11
19
null
null
0:00
meet the needs of what I call the next trillion core hours. The next wave of adoption is really in these both the training, the inference, and then pulling it all together and reinforcement learning and agentic applications. And so we're going to go through a few innovations and what they mean and why we're doing what ...
325
yVvLjWDmb0Y_12
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
12
19
null
null
0:00
It might be a network. And Kaslin will talk more about the networking side a bit later. But essentially, it's a way to describe in very rich detail the characteristics of the underlying hardware. So the workload then can express its requirements of the underlying hardware and get placed in a node or on hardware, on inf...
325
yVvLjWDmb0Y_13
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
13
19
null
null
0:00
And the second part is the decode, which is putting it all back together to return to the user. And it turns out that splitting these into two separate approaches is an important optimization and LMD is pushing this forward. Another aspect is caching. If you don't have to go all the way to the underlying hardware, but ...
325
yVvLjWDmb0Y_14
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
14
19
null
null
0:00
innovation moves on to the next step. So it's a repeatable pattern, a conveyor belt. One of my colleagues calls it a conga line, which I think is kind of funny. But really this is taking place in the workload API. And so some of the core features are scheduling multiple pods at the same time. And the value this brings ...
325
yVvLjWDmb0Y_15
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
15
19
null
null
0:00
becomes useful for every workload. So DRA is working for HPC workloads with our friends at schedmd now NVIDIA that are working on Slurm to make Slurm run better on Kubernetes because it knows more about the underlying CPUs. And now we're going to talk about networking and using some of the same underlying primitives. I...
325
yVvLjWDmb0Y_16
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
16
19
null
null
0:00
synchronized across multiple machines simultaneously are incredibly sensitive to network performance, so dragnets ability to assign dedicated high speed RDMA or TPU network paths can prevent network interference and dramatically improve data transfer speeds for your workloads. And I will mention the PM had me put this ...
325
yVvLjWDmb0Y_17
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
17
19
null
null
0:00
it does this at the load balancer level, while LMDB does it at the system level. So both of them can help make your workloads more performant. But one of them does it at the load balancer versus one of them does it at the system. LMDB has more stuff than that, but that's one way they're similar. We also released a blog...
325
yVvLjWDmb0Y_18
yVvLjWDmb0Y
Engineering the future of Kubernetes for AI at scale
Google
Google Cloud Tech
public
2026-06-25
null
null
As the primary workloads of the enterprise shift to AI agents, inference, and large-scale batch processing, architecting hardware-aware platforms with high elasticity becomes essential. Join this session to explore the cutting-edge landscape of Kubernetes. Learn how Google, as the top open source contributor to Kuberne...
[ "Kubernetes", "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
18
19
null
null
0:00
talked about them only about inference. If you want to talk about training, I'm probably going to head to my demo after this. Also, if you want to learn more about sandbox and the demo that I showed, follow me, and I'll take you there. But to review, we talked about AI applications and agents, which are instructions, t...
343
fkCTifAqVGg_0
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
0
22
null
null
0:00
Hi, everyone. Welcome I'm very excited to kick off the session today, where we'll talk about how to build a reliable agent from prototype to production. My name is alibek. I'm an engineering manager at booking.com leading the AI platform team. I'll be sharing the stage with amazing Googlers and my colleague Maria. They...
325
fkCTifAqVGg_1
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
1
22
null
null
0:00
facilitated over 6.8 billion guest arrivals in different accommodations. We are present globally, so it's a truly international company. We cover 175,000 destinations worldwide, and we have offices across 140\. We have 140 offices across 70 countries with $7,000 employees in our headquarters in Amsterdam, Netherlands. ...
325
fkCTifAqVGg_2
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
2
22
null
null
0:00
wanted to hand over to my colleague Maria to dive deeper into particular use cases. Thank you. Hi, everyone. I'm Maria, I'm a software engineer also in the AI platform at booking.com, and I'll build on what alibek just shared by showing you a few use cases with Gemini model. So the first, I want to focus on is called K...
325
fkCTifAqVGg_3
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
3
22
null
null
0:00
signal extraction. So a partner can go in and double check. Basically extract from comments and reviews. The common themes. So, for example, they can see that a lot of users, let's say, complain about the pool or the breakfast variety. And that way they know what to really focus on and what matters most for their custo...
325
fkCTifAqVGg_4
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
4
22
null
null
0:00
is a reserved capacity of tokens per minute. But that also isn't a bulletproof solution. So then we started looking into priority processing for multimodal use cases, for example, real estate reality. The real takeaway has been that we are trying to match the serving strategy with the production need, rather than just ...
325
fkCTifAqVGg_5
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
5
22
null
null
0:00
are going to cover four main topics in the next 30 minutes to help you understand how you can go about these challenges. First, we'll dive into how do you evaluate your agents before you put them into production utilizing various techniques. Then we're going to give some tips on how you can move from code to Cloud quic...
325
fkCTifAqVGg_6
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
6
22
null
null
0:00
to our evaluation systems. We can mainly split our methods of evaluating them into two components. First is response evaluation, where we disregard everything that happens within an agent and only take a look at the response. This is very similar to how you do any AI model evaluation. And the second is trajectory evalu...
325
fkCTifAqVGg_7
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
7
22
null
null
0:00
on. These are the automated metrics designed for agents specifically, some of which we inspire from AI evaluation like hallucination, groundedness, coherence scores that also utilizes LLMs judge. But you can also have custom automated metrics integrated into your systems. Now what I'm mainly going to focus on here is t...
325
fkCTifAqVGg_8
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
8
22
null
null
0:00
the right tools, that's OK. Then we have any order match. This is when you don't really mind. As long as the right tools are being called, the order is not really important for you. Think of us giving recommendations about London. You don't really care with which neighborhood you start as long as you give the correct r...
325
fkCTifAqVGg_9
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
9
22
null
null
0:00
aggregated metrics for the whole conversation. So by looking at that can decide whether if your agents are ready to go to production. And once you're there it is time to get to Cloud. And for that, I will hand over to Vanessa. We're going to be switching back and forth often, so don't need to clap every time, but we ap...
325
fkCTifAqVGg_10
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
10
22
null
null
0:00
order to bring that into Cloud SQL onto Google Cloud, the agent would suggest using Cloud SQL with Postgres. And also in order to make sure it's secure, it will also likely suggest that you need to make it a private instance along with Gemini Cloud Assist. We also have the application Design Center, and you can think o...
325
fkCTifAqVGg_11
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
11
22
null
null
0:00
make them work to do what you need. Now that we have the prerequisites, we make sure we have. We ask to design a Cloud architecture from our code base. So the agent looks through your entire code base, searches through it, tries to understand the components of it, and tries to figure out how they're connected and relat...
325
fkCTifAqVGg_12
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
12
22
null
null
0:00
can interact with our travel agent right there. It seems to work I think. Yeah it works. So one step further. So what I hope to show you with this demo is that if you have a process where you have an agent or you have an agent application, you built it locally, it runs well locally. And if you want to bring it to the C...
325
fkCTifAqVGg_13
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
13
22
null
null
0:00
you also need something like governance. So with teams building lots of agents and tools, you have a proliferation of them and you want to make sure not every agent that you have is able to talk to every other agent in your system. So for instance, in our travel app, we would like our travel app to be able to talk to f...
325
fkCTifAqVGg_14
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
14
22
null
null
0:00
let's get started with identity. We just released it today. Wear this is a solution where your agent will have a unique IAM at deployment time, and you don't have to manage it. It will have these privileges access provided by you. The advantage of having such solution, in the case of our demo, say that our concierge ag...
325
fkCTifAqVGg_15
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
15
22
null
null
0:00
you would prompt yourself over a long period of time to prevent from happening. And you would need to do rigorous tests. But if you were to place Model Armor in between, then most of these things are tackled on behalf of you. So let's see how that works in our demo. So here we have the UI of our application. We are gre...
325
fkCTifAqVGg_16
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
16
22
null
null
0:00
many of you have production applications. Almost all of you. How many of you have had production applications that have had no issues. Not a single hand up. That proves my point. So the thing is, the crux of the story is how do you figure out when something goes wrong if you don't have observability. And one of the thi...
325
fkCTifAqVGg_17
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
17
22
null
null
0:00
signals from the Postgres database, instead of signals from just that Cloud Run instance, et cetera. So with that, we were able to simplify our journey pretty quickly. And alongside with our agent app, we also have the Agent app dashboard, which has additional metrics that are useful to anybody who's running AI apps. S...
325
fkCTifAqVGg_18
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
18
22
null
null
0:00
traces for JNI look like. We'll come back to it at the end. So I promise you I'll show you this. So we only do one switch at the end to a live demo. So right now we have our application instrumented. It runs beautifully. There are no errors that I seem to be picking up, but there's still a very important question to an...
325
fkCTifAqVGg_19
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
19
22
null
null
0:00
AI model. And then finally the user gets the answer. Everything marked with JNI over here tells you that this has JNI information in the trace. So if we take a look at this see a whole bunch of metadata. You can see that we're using Gemini 2.5 Flash. The temperature that we used was 0.01. The number of tokens I used fo...
325
fkCTifAqVGg_20
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
20
22
null
null
0:00
find anything blah blah blah. So that gives you a 1 out of five. So just by using this data and running a SQL query, I was able to analyze the sentiment of every single user conversation. And that is very useful information for me. And lastly, I can also have a dashboard that I built to see the number of tokens that I'...
325
fkCTifAqVGg_21
fkCTifAqVGg
From prototype to production: 45 minutes to a reliable Gemini Enterprise Agent Platform agent
Google
Google Cloud Tech
public
2026-06-25
null
null
Simple agents are easy. Reliable agents are hard. Join us for a high-velocity master class on the production-first approach to Gemini Enterprise Agent Platform. We’re breaking down the complex transition from a basic prompt to a hardened enterprise tool. First, connect your agent by using Model Context Protocol (MCP) s...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution", "Security & Identity" ]
null
21
22
null
null
0:00
the guardrails, all the security requirements built in. And then you as a developer, just take that template using these MVP tools and customize it to run to your app. Then you make sure that your application, every single application in your organization is compliant. I'm going we have two minutes left, so I'm going t...
288
Z9Zz75pmOeg_0
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
0
16
null
null
0:00
Hope you're having a great Cloud ix so far. Really great to see you, everyone. My name is Niranjan, I'm a VP at Google working on our developer AI efforts. Joining me today is my colleague madhura, who's an engineering director on our team. And together we are going to talk about the agentic shift in software developme...
325
Z9Zz75pmOeg_1
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
1
16
null
null
0:00
through that in terms of various case studies about how we are handling this abundance of code and how it is putting pressure on systems like code review and testing and other places. All of the demos and examples that you are seeing, or all of the stats that we just presented, was all built using Gemini CLI and anti-g...
325
Z9Zz75pmOeg_2
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
2
16
null
null
0:00
flow time during review. Most of them are like, get out of my way. I'm trying to figure out what this is and what do I need to comment in terms of telling this author what needs to be fixed. So both author and reviewer. There was pushback and they weren't really ready. The users weren't really ready for it. There were ...
325
Z9Zz75pmOeg_3
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
3
16
null
null
0:00
experiment, active developer productivity for the Python cohort plummeted by 9% And we heard a lot of noise from users. They did not want this thing at all. So we pivoted away from this direction. And these are just a couple of stories to highlight. There was a lot of trial and error and trying things out for us to und...
325
Z9Zz75pmOeg_4
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
4
16
null
null
0:00
third anchor of our strategy is a risk assessor agent. This is scanning all of the in-flight changes for risks. And wherever we find it's a classifier, whichever in-flight changes have high risk. It adds a tag and says it's a high risk change. And all of these ones that are tagged with high risk require mandatory revie...
325
Z9Zz75pmOeg_5
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
5
16
null
null
0:00
are exposing these agents to Googlers. Let's look at how our agent sketching address the security issue. I've just left coded the server to View File content. The logic is there, but so is the bug which exposes internal source code. Let's see if our agents can fix this for me. Let me pull up the terminal to see what th...
325
Z9Zz75pmOeg_6
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
6
16
null
null
0:00
other thing that we were observing was developers just ended up taking longer time to get this agent to. Can you also do this. Can you also do that. And then it just slowed them down, instead of getting to that final output that caught all their design intents. So it was slowing them down. And then the last piece we sa...
325
Z9Zz75pmOeg_7
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
7
16
null
null
0:00
spectrum and testing agent is currently reading my code and existing tests and writing a specification document. The spec will serve as an interactive blueprint for my code, where the agent and I align on exactly how the code should behave and where it proactively services gaps like missing tests and ambiguous contract...
325
Z9Zz75pmOeg_8
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
8
16
null
null
0:00
to follow a different path than if you give it another time. And that's going to continue playing out. Depending on how complex your system is, how large your code base is how many tools they have available, and what all different paths that these agents can take. The other thing that is still evolving, and we've been ...
325
Z9Zz75pmOeg_9
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
9
16
null
null
0:00
that is happening. So we can have access to it at an aggregate. And of course, the agent output, we're also annotating agent output with additional signals. So we can try to make sense of what is this output good or not for however we are trying to match the task and good output. And at an aggregated trajectory level, ...
325
Z9Zz75pmOeg_10
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
10
16
null
null
0:00
on the product experience. There is a left side tab is showing all of the substeps the agent is taking to complete the task. You'll see blue lines middle tab that's showing the duration of each of these events, the substeps. And then there is another tab on the right that is going to show for each of the steps what all...
325
Z9Zz75pmOeg_11
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
11
16
null
null
0:00
any combinations of there. Between that being said, when you are a large company like Google with 100,000 odd developers, you need to absorb the pace of change in a way that actually makes sense for the organization, for the variety of different problem domains that they're going after. We have ML engineers working on ...
325
Z9Zz75pmOeg_12
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
12
16
null
null
0:00
of really smart people in the company, and many of them are extremely enthusiastic about AI. Some of them are at the bleeding edge, keeping abreast with the latest technology developments, whether it be a new model or a new harness or what have you. And so they tend to develop a good amount of gut feel and gut intuitio...
325
Z9Zz75pmOeg_13
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
13
16
null
null
0:00
that, we build and share a lot of dashboards across the company. At this point, most major companies have some sort of an internal token use dashboard. We have similar ones, but our goal is to really help engineers understand whether tokens are being spent and use them effectively for the right kind of meaningful tasks...
325
Z9Zz75pmOeg_14
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
14
16
null
null
0:00
so you can do some kind of Hill climbing on your own tasks as a result. We also have this category that we are calling TM agents. And madura showed you a couple of examples of the code review one, and there's a lot of tribal knowledge that resides within a team. Best practices, coding standards, certain ways of deployi...
325
Z9Zz75pmOeg_15
Z9Zz75pmOeg
Navigate the agentic shift in software development with Google
Google
Google Cloud Tech
public
2026-06-25
null
null
The future of software development is AI native. At Google, we integrate AI deeply into every stage of the software development life cycle (SDLC), moving beyond code assistance to create a collaborative environment between developers and intelligent agents. This session will explore how Google built an AI-native SDLC, ...
[ "AI / Machine Learning", "Agents & Autonomous Execution", "Cloud Architecture" ]
null
15
16
null
null
0:00
tuned the sandbox. They can be pretty powerful, and in fact, they can actually help you reduce your risk and technical debt. Write code is not just an asset, it can be a liability. The more code you have, you're likely carrying more risk. And if you actually use agents correctly, if you use agents in a way that are try...
176
LHcjN11nNPU_0
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
0
16
null
null
0:00
Good morning, everyone. Welcome to our session scale AI agents in production. I'm Ryan ismert. I'm a product manager on the Gemini app Enterprise Agent platform team, and I'll be joined here today by my colleague Ilya and by two of our favorite customers, Praveen from Comcast and Tejal from Palo Alto Networks. We have ...
325
LHcjN11nNPU_1
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
1
16
null
null
0:00
your employees and your organizations to do things that were never before possible. Investments in your organizations and agents are being driven by real benefits. So how do we deploy and scale agents to maximize this value. As we announce this week, Gemini Enterprise Agent platform is our offering to help you realize ...
325
LHcjN11nNPU_2
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
2
16
null
null
0:00
using identity, agent, gateway, and agent registry. And finally, our optimized pillar where we give you tools for agent simulation, evaluation, and observability. Ellie will touch on these as well in his demonstration. Let's talk for a minute about some of the recent improvements and upcoming features for our agent run...
325
LHcjN11nNPU_3
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
3
16
null
null
0:00
automation. It offers network egress controls and support for agent identity. Additionally, we offer additional safety and human in the loop features through integration with our computer use tool, but you can use any computer use model that you desire. Finally, if you need extra flexibility, we offer bring your own co...
325
LHcjN11nNPU_4
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
4
16
null
null
0:00
we iterate more and more into this phase. Finally, when we reach a certain level of confidence, we say, let's actually try to deploy to a production environment. So we start the deployment process CI/CD finally, we're in production. And then after we're in production, we start observing how our users are using our agen...
325
LHcjN11nNPU_5
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
5
16
null
null
0:00
to enterprise for example. All that is possible. So let's have a look through the demo on how it looks like. OK So I'm going to switch over. And for this demo, we are going to use Gemini CLI. Of course, let's try to see how it works by building such agent. We're going to ask a simple question. This is a demo, but we're...
325
LHcjN11nNPU_6
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
6
16
null
null
0:00
I'm going to put in YOLO mode so that we can actually go fast on this. But you can see that the first phase will be about and designing aspect so that you can approve, the plan that the agent will carry on. And, you will have example use cases, constraints, tools require success criteria. And I expect you in a real sce...
325
LHcjN11nNPU_7
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
7
16
null
null
0:00
to ensure robustness. Yeah so super super cool. Now it's failing. But it might take some time and depending different iterations to do that. Let's actually stop it. And let's say I'm OK with it. I want to deploy it to agent runtime. I'm OK with it. I want to deploy to agent runtime. What will happen here is that the ag...
325
LHcjN11nNPU_8
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
8
16
null
null
0:00
do here, we will explore our environment, our deployed agent. It will look at the view. We have a specific BigQuery view where we are storing all the traffic that is going through production through the agent, and it's starting running analysis on top of the data we are storing in BigQuery, which is so cool. So here is...
325
LHcjN11nNPU_9
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
9
16
null
null
0:00
our customer experience with the power of AI. So you all have seen Xfinity assistant. This is our digital chatbot where customers can come in. They can get the real time information about their services, make payments, troubleshoot their services. This is all contained within the digital platform and our engineering. O...
325
LHcjN11nNPU_10
LHcjN11nNPU
Scale AI agents in production
Google
Google Cloud Tech
public
2026-06-25
null
null
Take agents from prototype to scaled production with Google’s modular, generally available agent platform services. Join this session to learn how to scale agents using our newly revamped runtime, which now offers fast cold starts and provisioning, personalization with memory bank and sessions services, and secure sand...
[ "AI / Machine Learning", "Gemini / Google AI", "Agents & Autonomous Execution" ]
null
10
16
null
null
0:00
them here. We will go through this in detail, but I'm not here to say everything is awesome. Every one of you in this forum may have faced any of these challenges, and we good news is we have solutions for all of them. We work closely with Google Cloud team to brainstorm ideas and come up with solutions, not digging de...
325