Loading summary
Transcript34 lines
- [00:00]
Kaslyn Fields
Hello and welcome to the Kubernetes Podcast from Google. I'm your host Kaslyn Fields.
- [00:06]
Abdel Sighiwar
And I am Abdel Sighiwar.
- [00:18]
Kaslyn Fields
We're back with a new episode. After a long pause, we hope you missed us. This week we speak to Amy Wei. Amy is a Senior Solutions Engineer at Hashicorp and was the winner of last year's GKE Turns 10 hack. It was Amy's first time entering a hackathon and she ended up bringing the prize home with a cart to Kitchen AI Assistant. But first, let's get to the news.
- [00:45]
Abdel Sighiwar
Kubernetes 1.36, codename Haru is here. The new version introduced 70 enhancements. Of those, 18 have graduated to stable, 25 are entering beta and 25 have graduated to alpha. We will be bringing the release lead on the podcast to talk more about this soon. Stay tuned for that episode.
- [01:03]
Kaslyn Fields
LLMD was donated to the CNCF as a sandbox project announced at Kubecon EU 2026. The project aims to provide the industry with well lit paths for efficient inference on Kubernetes. The project was founded by Google Cloud, Red hat, IBM Research, CoreWeave and Nvidia, aiming to unite around a clear industry defining vision any model, any accelerator, any cloud.
- [01:30]
Abdel Sighiwar
The CNCF introduced the CARE program or Certification Advancement and Recertification Experience, a new policy to make maintaining certifications easier. Since most certifications expire after a while, typically two years, the CARE program addresses this by allowing advanced certifications to extend the validity of related foundational credentials. For example, the Kubernetes and Cloud Native Associate KCNA will automatically renew when a practitioner passes or recertifies a on the certified Kubernetes Administrator CKA or Kubernetes Certified Application Developer CKAD. The CARE program will be fully implemented in June 2026 and will be effective as of January 1, 2027.
- [02:10]
Kaslyn Fields
Agones moves to the CNCF, the open source platform for scaling and orchestrating dedicated game servers on Kubernetes celebrates its official transition to the Cloud Native Computing Foundation. Originally founded as a co development between Google and Ubisoft in 2017, the project is now being donated by Google to the CNCF at the sandbox level to foster a community owned and governed future. Congratulations to the project team and that's the news.
- [02:42]
I'm Kaslyn Fields of the Kubernetes Podcast from Google and I'm here at Kubecon Cloud Native Con North America 2025 in Atlanta where I am speaking with Amy. Would you introduce yourself please?
- [02:52]
Amy Wei
Amy yes, I'd love to. I am Amy Wei. I'm a solutions engineer at Hashicorp.
- [02:58]
Kaslyn Fields
We've been so excited to have you all day. We've been doing interviews with you kind of all day and for the whole week, really the last few days, because you are our global winner of our GKE 10 year anniversary hackathon. So would you tell me a little bit about your experience with the hackathon and what you built?
- [03:16]
Amy Wei
Yeah, this was my first hackathon and glad to do it with Google on gke. And I built a AI application that is an online grocery store and I built a AI assistant that takes in the user's shopping cart and suggests recipes based on the ingredients. And that makes shopping for groceries and meal planning a bit easier.
- [03:43]
Kaslyn Fields
And to give folks a little background on this hackathon, the challenge was we have a couple of sample applications, Online Boutique and Bank of Anthos. And we've used those sample applications in like keynotes and presentations for years. And when we were thinking about how to celebrate 10 years of GKE and developers, we were thinking, you know, we want to get folks hands on, especially in this current era of AI where everything is new to everyone. And so we wanted to give folks an opportunity to see what AI means for gke and we wanted to really focus on the agents and what you can do with AI and the creativity of thinking about that new area. So we didn't want folks to focus on building the coolest kubernetes infrastructure. We know people can do that. We want to see what kinds of agentic solutions and ideas people can come up with. And thinking about it kind of in an enterprise use case businesses, you have an existing application and now you're getting the word from on high that you gotta put AI into it somehow. So be creative with it and have some fun with it and put some money behind it. So then there's an incentive to explore it. And so that was the challenge to take an existing application and to build onto it. And how did you approach that challenge?
- [05:11]
Amy Wei
Yeah, this was a really fun challenge. I think building on something existing is a very realistic challenge because very rarely, unless you're a startup, you're building something from scratch, you're usually looking at other people's code base and trying to figure out what the heck was built and what was the logic. And that was also kind of challenging for me because I was a solo developer. So I had to balance the programming aspect with the presentation as well as the storytelling aspect as well. But AI has made that much, much easier and Much more approachable for myself because I could have AI explain the code base to me and explain the new protocols like for example, A to A for Agent to Agent as well as ADK Toolkit and have AI explain that to me and I can ask all kinds of dumb questions and not feel super self conscious about it. So yeah, it's very user friendly.
- [06:09]
Kaslyn Fields
This is your first hackathon solo developer and about how long did you spend building your solution?
- [06:16]
Amy Wei
Yeah, it was on and off a little over a month. So took some time to come up with the idea of the grocery shopping experience and then the kind of the programming and just getting everything fleshed out and then the last portion, probably about a week was on the presentation as well as just putting all of the content together and you know, checking off the list to make sure that I met the requirements.
- [06:42]
Kaslyn Fields
Yeah, we had quite a few requirements for this hackathon, which I think is an important part of the design. We did encourage everyone doing this hackathon to use Vibe coding to create their solutions, which I think is very interesting. Probably scary to a lot of people because I think that's going to be the way that hackathons go now is you're definitely going to get a lot of Vibe coded solutions, so you might as well prepare for it. So in our case we really wanted to encourage people to try out that new technology to see what it could do, to see how it could help you build something that works and see your idea in action in a pretty short time frame. So how did that go for you? What kind of Vibe coding tools did you use and what was that experience like?
- [07:25]
Amy Wei
Yeah, I used VS code with Copilot and then I tried the Gemini model. I think I tried the cloud model as well and then also the Gemini CLI for just like general questions. Let's see, what else did I use NCP servers for the agent to look up documentation because I think maybe because the protocols are pretty new. So the agent did have quite a lot of hallucinations, but looking up the documentation like the developer docs through the NCP server really helped improve some of that.
- [08:00]
Kaslyn Fields
That's a lot of different AI tools. Just going into your process of building the application, like the process around that, not the application itself, which also used AI. And you gave us a little overview earlier of what the application did, but could you speak a little bit to what areas within the application you thought AI was a good fit for? And where you used was fun to
- [08:24]
Amy Wei
think about where AI could fit. For me, the main functionality was the Recipe generation from the, you know, a user's card. So the recipe, like creating the recipe as well as the image that goes with the recipe was done with the Gemini model as well as the image and model. Because I feel like with these large language models there's still some creativity. Like the responses are not gonna be the same every time. So there's a little bit of leeway and I think recipes is a place where it's okay to have that variation. That could sometimes be a good thing. So I thought that was a good fit.
- [09:04]
Kaslyn Fields
Yeah, that's a really good point. I hear a lot of like summary use cases. It's really good at taking a bunch of information and summarizing it. I love the idea of it's non deterministic, it's not gonna do the same thing every time. So make use of that in the design for sure. And so we talked a little bit about Vibe coding and this being your first hackathon and your idea. I wanted to also talk about the architecture that you chose, the different tools that you use, and the requirements of the hackathon. I think in a world where people are vibe coding their solutions to hackathons, one way to manage the quality of the submissions that you get is to have a variety of different types of requirements. And for this one, we required that folks had an architecture diagram, a GitHub repo with their code video, so a variety of different requirements. So how did you approach those requirements and the architecture that you built?
- [10:05]
Amy Wei
Yeah, now it makes much more sense given the requirements, and it's to make it a little more difficult and not rely on everything with AI use. I think for me, just understanding the requirements and making sure that I had enough time to tackle each other requirements was really important. I think some of the requirements being like using the ATA or adk, so understanding what those protocols and the frameworks are. And then I think at the end when I was doing the final check to make sure that my submission met all the requirements, I fed the code base as well as some of the content from the presentation. And then I asked AI to basically do a check against the requirements. So that was really useful.
- [10:56]
Kaslyn Fields
I think that's really common. From what I saw in a lot of the submissions, it was like, well, they clearly checked off the box here and here and here. Also in the code for a lot of the submissions. A lot of people didn't actually check off those boxes that they said they did in the, in the. Perhaps AI generated documentation. But I think it's. And it's Important to also look at the ways that you can use AI to complete the other pieces of the hackathon and how that's different in the different pieces. So like Vibe coding, you might be in a different mentality than when you're trying to guide the AI to help you write documentation. Or some folks also used AI to help build their videos. Like they used it to create B roll or as an AI voice to do the voiceover for them. There's all sorts of ways that used AI.
- [11:48]
Amy Wei
Yeah, you can get really creative with it and it's really become a helpful assistant or almost like a team member, which made this possible because I was the solo developer, I think maybe without AI this would have taken much, much longer and my scope would have been reduced significantly.
- [12:08]
Kaslyn Fields
And I think that creativity of thinking about how to use AI is the benefit of doing this as a hackathon. A lot of folks, you're working in your day job and you're very focused on the quality of what you're building and building an enterprise grade solution that's going to be maintainable. And a lot of people worry about how AI fits into something that's long term like that. Whereas with a hackathon and having all these different requirements and encouraging the use of Vibe coding, we're saying just play around with it, see what works. And we wanted to show off the solutions and the winners because a lot of folks are looking for use cases. Like, I'm being asked to do this. Maybe I don't really want to do this in my day to day job, but how can you make it fun and how can you be creative with it to find use cases that are actually good? So we really love that your solution was a very complete view of what you were trying to do. You defined your like it was a grocery store and you were looking at the rest of the items in the person's cart to help inform the recipes as well as if you didn't have something in stock. Like how often are those coming up in the recipes that people are getting for their items? Or how often are people trying to. I mean, I'm sure businesses do that already. Like how often are people looking at that page when it's out of stock? But you told a really good story of like both the business side and the user side and then implemented the whole thing really effectively.
- [13:41]
Amy Wei
Thank you. Yeah, this was again just a really fun project because cooking and grocery shopping is a pretty big part of my life, so this was fun.
- [13:51]
Kaslyn Fields
And one more thing I want to talk about it is the GKE 10 Year Anniversary Hackathon. But we were focusing on the agents and giving people an opportunity to play with the new AI tools. But how much did you know about Kubernetes going into it and how much did you learn through the process?
- [14:08]
Amy Wei
Oh boy. Well, I work at Hashicorp, so one of our products is Terraform. So Terraform, you can use it to spin up a lot of different services, including gke. So I had a head start in that. And using Terraform has made the creation of the cluster much, much easier. And then we also have other integrations with Kubernetes, but more on the security side. So there's some experience. But I definitely learned a whole lot more through the hackathon, especially like on the application development side as well as, you know, just incorporating AI and just seeing how the protocols actually worked through a hands on experience was really valuable because that made it more tangible for me.
- [14:54]
Kaslyn Fields
I love to hear it. And one thing I didn't ask you earlier, but I do want to ask about is how did you find out about the hackathon and decide to go for it?
- [15:05]
Amy Wei
Found it online, I think it was through Reddit.
- [15:08]
Kaslyn Fields
My team made that post. Yeah, my team made that post. So I'm very excited about that. So to kind of wrap things up for other folks out there who might be interested in going for hackathons and maybe thinking about gaming the system with writing their own agents or doing vibe coding, I mean, we encouraged it here. What advice would you have for folks who might want to try doing something like this?
- [15:35]
Amy Wei
Yeah, definitely dive in. And then there's all kinds of tools out there. Take advantage of those tools and all the free credits and see what you can build.
- [15:46]
Kaslyn Fields
Yeah, wonderful. Thank you so much, Amy.
- [15:48]
Amy Wei
Thank you.
- [15:51]
Kaslyn Fields
That brings us to the end of another episode. If you enjoyed the show, please help us spread the word and tell a friend. If you have any feedback for us, you can find us on Social Media KubernetesPod or or reach us by email at Kubernetes podcastoogle.com youm can also check out the website at kubernetespodcast.com where you'll find transcripts, show notes and links. To subscribe, please consider rating us in your podcast player so we can help more people find and enjoy the show. Thanks for listening and we'll see you next time.