Google Cloud Serverless: Accelerate your AI

Key Takeaways

[Music] I'm here at Google Cloud next with Sarah and she's going to talk to us about the talk she did about accelerating your AI with serverus Sarah thank you so much for being here with us and chatting with us yeah thank you thank you for having me of course it's a pleasure uh can you tell me a little bit about yourself what do you do here at Google and a little bit about your talk yeah sure I'm a developer relations engineer at Google on the serverless team yes wonderful and your talk was accelerate your AI with serverless so it was basically an overview of how Google Cloud does serverless AKA Cloud run and also had a a three demos that show how you can use vertex AI with Cloud run oh wonderful okay so in these three demos can you walk me through what were these demos and why did you choose them specifically yeah so I'll start with the why uh because I wanted to show different contexts where you could use serverless or Cloud run where you could use a cloud run service with uh with vertex AI or with Gemini so the first demo was a Gemini demo which showed how to do a cloudon service with a Gemini power uh Gemini powered chatbot then the second one it was called um very similar to the first demo but it used function calling so basically LS can seem Frozen in Time based on their training data for example if I ask Gemini hey what is the uh current weather in Seattle it's going to say I don't know because uh it's training data ended April of 2023 right but with function calling you could go to like for example a cloud run service and find out what is the current weather in Seattle which I hardcoded as 40° and rainy because that is the yeah ask me how I feel about Seattle and so that was demo number two so you make the call on Gemini's behalf and then demo three was using a different type of cloud run resource there server there's um Cloud run services but there's also Cloud run jobs and a a cloud run job just runs a container or the code to completion and so I showed how to use different non- Gemini specific uh vertex AI apis in that Deo wonderful okay and just to give Folks at home maybe a little bit of the background what is serverless and what are the benefits of it yeah sure so uh server lless is where I would think of it as like you Outsource your Sr or your devops to Google so where you get to focus on your application development and uh your performance monitoring and we will take care of Autos scaling up if you have more traffic or scaling down like if you're like me you're experimenting you're prototyping you put you spin up an instance but then you don't you can just forget about it because after 15 minutes of idle time uh Cloud will spin down that instance based on traffic oh got it okay okay so then uh in terms of using Cloud run you know a lot of people ask where should I run my code yeah yeah so so usually I always start with okay where are you where are you starting from like are you comfortable with containers or not if you're let's say like you you just have your code and that was my first and second demo uh demos 1 and two where I use build packs to show that you don't have to have a Docker file we will containerize your code for you so you can just bring your code uh for I think there's seven languages that are supported by Bill PX and demo number three I actually use a Docker file because I had to install a third party or a third um third party binary um ffmpeg for that for that specific demo but usually yeah I always like like are you comfortable with kubernetes if you are check out gke okay wonderful okay so if uh folks want to watch maybe other sessions on servus that have been happening at next what would you recommend yeah so there's uh there's what's new in Cloud run where they demoed this morning in the keyot the uh app canvas where you can describe your architecture in actal language and uh it will produce not only a diagram um of the your architecture but you can actually hit deploy and it will actually deploy that yeah so that was uh my teammate vza so if you want to see this live go to his session and uh staring the PM for uh Cloud run it's uh it was called um generative AI using Lang chain with Cloud run but yeah it was it was pretty cool he asked I want a cloud run service with a cloud SQL and then it it generates a scaffold or it shows you what the architecture will look like then he says give me a cloud run uh a cloud sequel um Cloud sequel give me a uh I want to use a Gemini and then it wires up gemini or vertex AI integration and then you hit deploy and it deploys it for you yes okay I did the manual way cuz uh I wanted to show step by step uh what is going on when you're writing code and I also want to just call out that um everything I demoed in my talk I have a step by step by step guide in the Cod laabs so I told my audience do not take notes take a screenshot of these QR codes everything is already taken care of you go to the code laab you can start from scratch wonderful okay and we'll make sure to link those code labs in the description as well so all of you can check those out and okay I'm going to throw you a little bit of a curveball here if uh developers want want to get started you know seamlessly integrate this into their workflow what can they do can you tell me in 60 seconds or less if you want to get started with Cloud run you can go to the cloud console and from there you will see how you can deploy an already uh uh created container image of the the hello world Happy dancing unicorn that is what I call it and you can just go from there and get a feel of the auto scaling I would say yes start there in the cloud console get an idea of the auto scaling scaling up and down the knobs and dials that you can find tun for yourself and then you can do the source based deployments uh it's kind of like the next step in our documentation so that you can get your own code uh deployed and then you can go from there from like continuous deployment wherever you have your get repo and so forth wonderful thank you so much for chatting with us Sarah and have a great rest of your next thank you for having me thank you [Music] oh [Music]