النص 🔗
Jatin: everyone. Thank you, Zishan, first of all, for inviting me and having me here. Appreciate it and means a lot. I'm Jatin, ⁓ developer advocate at AWS. Before ⁓ landing this role with AWS as a developer advocate, I was in Japan building applications, building cloud infra ⁓ on AWS and due to this I was also a community builder. Where my main outcome was like to give produce content, ⁓ whatever I learn and share with the community because learning ⁓ becomes two X, three X when you do it with community. So for folks who are looking to be a developer advocate or s or anybody who wants to learn, share out in the community. That's what I can say. So it the answer would be depends. Okay. I'll tell you how it depends. Because now you have everything right now at the edge of AI. AI. Right. Right? So it depends whether you one, if you are a builder at heart, two, if you're gonna be curious. These are the two factors which I feel ⁓ if you have these two factors, then your day-to-day, like for me, my day-to-day has never changed. Right. I still I still build code. I still push ⁓ new features for EKS. I still write ⁓ you know, new agents because I'm a builder at heart. Right. Why I still do this? Not because it's my job, because I want to teach people, I want to aware people what the world is going on and how it can make your life easy as a developer. Right. So you need to be builder at heart and you need to be curious. It's never gonna be changing if you're coming from let's say a product marketing, which you asked, right? I'm assuming it's gonna be less technical. But if you have that curiosity and you are a builder at heart, with AI, you can leverage these two factors to build new things. So it doesn't matter in that way. It only matters with these two factors. Right. ⁓ Exactly. Even with AI, now with these two factors, there's one more point which I just came into my mind. Empathy. Right. Do you empathize with your developers? Do you ⁓ empathize like what if they're gonna be building something, how it's gonna be solving their problem? Right. That's what we also try at AWS to empathize with developers, be where developers are. Right. So if you're coming from anywhere right now, the barriers have been removed. Yeah yeah. The gatekeepers are removed. Now it's your job to be curious, to be builder at heart, yeah, and show empathy towards your audience. That's what it matters, at least to me. Very interesting. very interesting question. And we keep coming in this deprivable space this question. Somebody who can core and somebody who cannot. Exactly. With AI, like I mentioned, now that barrier is has dissolved. Right, right, right. So it's still, you know, what matters right now, whether somebody can code and somebody cannot, will be the experience. You know, the human intuition and the experience. Because even if you cannot write code, if you have that experience of okay, this was my product when it was shipped, and this was the problem my developers face, that experience, that human intuition can bring a change. Right. But if you having said that, But if you're on that side where you can code, you know, ⁓ then then you are you're better con can connect with your audience because they can understand, okay, this is where I was facing. I when I was coding, this is the exact problem I faced. Exactly. But with no offense and no disrespect to senior folks or to the freshers who are coming in who doesn't have experience, who leverage AI, you know, to write code, which is not wrong. It is not wrong. Again, the difference between senior folks and the new freshers would be the experience and the human intuition. Yeah. So good. ⁓ you don't have to write code that way which we were used to write. But now, at least don't just ⁓ overload or offload everything to AI. Exactly. Have your human intuition with you. And that's where, you know, even if you don't write code, even if you're writing code with AI, that will help you. This is exactly. For sure. Hmm. So what I feel is ⁓ again, because I have been through that journey of being ⁓ infra guy, DevOps guy, I know what possibilities or what problems or possibilities a developer will face. Right. Right? That's the my intuition. That's where my experience is coming into the picture. Because I have experienced that, I have lived that. So where I feel that when a developer is building his application, his or her application, right? And that's where the production issue will arise. And that's where the key components of because building an application involves complexity. Complexity is never going anywhere. Right. So when I feel like this is the part of the application which comes to the complexity, which will bring in complexity. I won't stress on that point. And where I feel, okay, this is mentioned in the docs, this is just a surface level. Then I would love to like, okay, let the docs do their part. But I will do my part in helping the developer journey in using something to reduce their complexity of the code. That's where I want to go deep. Hmm. So yeah. ⁓ I think it's both. Because ⁓ it's all about the experience of a developer, right? Some something which was working earlier and now it doesn't work, that's a no no for ex from an experience point of view. Yeah, right. And if something is coming new and developer doesn't know how to use it, right? How to ⁓ simplify their development journey, that's again a no-no from an experience point of view. It's both. Now depends on chasing, like chasing next or ⁓ going back. I think it's a part of the role to fix. At AWS, we always have this mindset as a developer relations. Be where developers are. If developers are using something which was working, which was like a technology old ten years old technology like Kubernetes, right? Right. And something is not working, like our sample codes are not working, we will fix it. We have to fix it, right? But again, like let's say, now the things now we are aiming to improve
Load full transcript
productivity, efficiency using AI. That's what all that hype is about, right? Everybody's talking about that. With efficiency. So improving productivity would be I would ⁓ so improving productivity is one construct. That's where I want to chase by use leveraging AI. So it is both. As a DevRil, if you want to be a part of DevRail, you have to do both. You cannot ignore one or the other. Yeah. You have to do both. And that's what DevRil job makes us fun. Yeah. I mean, ⁓ How when you were doing your ⁓ you said you were a community manager, right? Mm-hmm. You work backwards from the community. Yeah. Right? That's where we work backwards. If something is not working, it's fix our thing that's how it is. Improving the developer experience throughout the journey. It's a close loop. Yeah. You have to close the loop always. Chase at the same time, fix whatever is working. Yeah. It's a loop. Okay. Right. Okay. I get it. So let's say you met me, you are in that 15-minute tutorial or a workshop, right? And you have no idea of that AAA's DevOps agent. Uh-huh. Right. Again, from an empathy point of view, as a developer, what would you expect from me? What is this DevOps agent? What problem does it solve? Following up, why do we need this DevOps agent? And a basic premise is. How I can leverage this or what it is being used for so that I can now leverage this in my in my day-to-day you know developer cycle or my operation. How simply I can help it. That's all can be fitted for me, can be fit in those 15 minutes. You know? Yeah. How, what, why, where I can use it. Again, and how just to give me a some demo. So that I have that confidence to use it. Theoretical knowledge is very nice, very good. But until you see it, we humans are very like, you know, we are very simple. Right? We can read it. Yeah. But we won't understand it until we see it visually. That's that's our simplicity, you know. Some people may argue this is our complexity. But I feel we are very simple in that way. We can read it. But I don't understand it when you're you cannot show me. That's all those things will fit in that. Four 15 minutes, right? But when you say what fits in the prod now, now when it fits in the prod, now you need to think up beyond those, you know. What are the how do you relate beyond those premises of how do you relate this? Let's say, for example, you mentioned DevOps agent, how DevOps agent will help me in my day-to-day operation. Right. Right. Is it just gonna be helping me to fix some incidents? But now how can I integrate with Slack? Right. Or how can I integrate with my existing tool set which improve my productivity and efficiency? Simplicity. Simplicity, right? Now this is where you go beyond just the feature set. Right. For example, for example, DevOps agent integrates with ⁓ tool observability tools like New Relic, Datadoc, right? Now, as an engineer, When something comes in your AWS accounts, do you want to go manually and fix issues? Right? But in 15 minutes, I
can show you how do you manually investigate. Right? But when you're coming into the prod, now you will integrate DevOps engine with your tools and have an automatic investigation through webhooks. That's just I'm telling you a use case. Yeah, right. Of course. And this can happen through Slack. Like notification comes and it goes to Webhook and Webhook just in ⁓ throws DevOps agent, please investigate. And then DevOps agent gives you the return back through the slack. This is how you now you make your life easier when you're working in a production environment in your enterprise level. Enterprise level, right? Right. So that fits in now in the ⁓ production level. But in fifteen minutes, what I can just show you is let's say an E case cluster, there's an error. Right. Right. How do you manually investigate? To show you the power. But the premises will come, how do you integrate for your enterprise level in your production? That's where the difference comes. And that's where I draw a line. Right. Familiarity. Yeah, familiarity. That's a huge. That's a huge number. For sure. For sure. ⁓ ⁓ which I wrote yesterday. Okay. Now this is something now we are going towards. You know, I don't have a definite answer to this to be honest. As a Devril, this is what I always ⁓ tell people. ⁓ be authentic. If you don't know, you don't know. Yeah, exactly. Right? You don't have to ⁓ bluff about it. At this moment, this is the topic of the industry which I feel there's no solution to this. Right. We are building new forms of ⁓ augmented tools like skills. Like you know, you're giving the directions. ⁓ recently AWS brought something called as agent toolkit. So the idea remains the same, similar to we have skills in that as well. So idea is model does models knowledge what whether you are agents or whatever, they rely on the brain as the model, right? They rely on outdated information. The model's information is as good if it's not outdated. Exactly. Right. New features are being shipped at the unprecedented speed now. How do you keep up with this model training, right? Exactly. They knew to get they have to give new versions of the model, right? Because training has to happen for the new data. That's where now industry is trying to close the loop. Like AWS Broadagent toolkit. Right. We have now an MCP server, which whenever a new feature comes, all our APIs are there. It's also talks with the documentation, right? We have a skills all curated, ⁓ checked, ⁓ tested by the product teams, AWS teams, created by AWS. So all to ensure the efficiency and correctness of agent. That's agent toolkit. But when you ask me this question, like is there a Definite answer, like how do you ensure correctness? These are the different ways which we can do. Skills, agent toolkit. Are we there yet? I think no. No. You know, not yet. Not yet. ⁓ but I think this is where we this would be a new idea for our blogs, you know, if we are collaborating next time. Yeah. How do we ensure the efficiency and correctness of the agent? Until now, the problem was we were having these different hacks. How to run agent in an isolation environment. ⁓ you know, have startups like, you know, a user is thinking something, he comes back, they it's a idle, you know, the agent sitting idle. How do you suspend and resume again the agent work environment and near startups? This is all now being solved. Like two days back, micro BMC is solving this problem. And and let's just say it's just not about agents anymore. It's any kind of an application. You know, we used to have hacks like containers goes there. Different isolation. But containers in ⁓ always share a kernel behind the scenes. Kunnels. They share kernel. They do. And if a C V is on the cur on the ⁓ kernel level, your entire code is bleached for all the containers. That's one. So this is being now managed as a la AW service lambda microbeams. So we can think like how I feel is ⁓ the infra part is now being sold. Now the next part is the Correctness. We are not there yet. But I think now industry would be now working on that. Right, right. There's again a new problem, token maxing. Yeah. Exactly, right? Yeah, I saw that. I d I saw your poster Twitter, you token maxing with opus. Exactly. This is like again now industry is moving sick. So technology started. Everybody we are now creating awareness. Like how do you use agents for your work? Efficiently. Efficiently, you know, improve your productivity. Now the engineering problems are arise comes. Token maxim. How do you reduce the cost? How do you reduce the cost? You know, how do you improve the efficiency? One engineering problem for infra, solved. Right? But now the other problems are coming. So we are in a very interesting space. So whoever looking at your episode as an engineer, you are not gonna be replaced. Your engineering problems are just getting more interesting. And more complex. So you're in a different space now. And it's a exactly. You have all the tools, how to solve them. Now you have a newer set of problems. So people are kind of scared these days, right? That okay, we're moving ahead of this question, but I just want to say people are not scared that okay, as a developer or as an engineer. I may not be relevant. Right? I'm not saying I'm not using ⁓ deliberately, I'm not using the word replace, but I'm choosing the word ⁓ relevant. No, I would say you are very relevant because you would be the one to drive or use those agents, models, these infra. Your engineering problems, your engineering mindset has to be changed. Exactly, exactly. More than upgrade, ⁓ one of my senior leader recently, ⁓ I had this amazing conversation with him and he mentioned an engineer in this time, you know, when we are talking, is that ⁓ an relevant engineer, sorry, let me reframe relevant engineer is that engineer in this time who embraces change. Flexible to adapt to this newer technology. Who who you know? Just like, okay, earlier my engineering problems were different. Now they are becoming more different. More different. They are different now, you know, different than previous. That is the engineer who's gonna be thriving in this AIH. That's ⁓ that's what my lesson from that conversation happened, and it and it stuck with me. You know? ⁓ so whether you're a community manager coming from product marketing, mm-hmm, a DevRel or a veteran engineer or an you know, a college grad, fresh grad, mm-hmm, an engineer who's gonna love embracing the change and is flexible to adapt is the one who's gonna thrive. ⁓ yeah. Lambda microbium fanboard. Okay. For sure. ⁓ I think isolation is sold now. You know, it's a managed primitive from AWS. ⁓ proving it right will be the next frontier for all the engineers. How do you ensure correctness? How do you reduce hallucinations? That's where the different tool sets are coming up. We are not there yet, I feel like. And ⁓ that's where, you know, how again. These all these conversations we are having around, you know, the models, infra, they are they are to be increasing your productivity efficiency. But now, as an engineer, how do you use it efficiently is also one more problem, which you need to solve and no AI agent or model can solve. So be creative, be curious, and be hungry on how you can solve these problems. You know. Difficult and different. That's where ⁓ I think. People will thrive. Yeah. ⁓ I see you are very active on Twitter. So you do amazing things on Twitter. I need to learn that from you. What is T is for the audience? For sure. That's something we can work on it together, you know? Like find some ways. Or next episode can be for our audience that okay, this is how we are ensuring not hundred percent, but at least seventy to eighty percent correctness for our agents. Right. That's where it can be. Yeah, I can I totally agree with that. Thank you, Dishwan, for having me here. And it was lovely talking to you. ⁓ let's do this again. And you know, for the audience, for our developers. Thank you.