Cloud AWS re:Invent Coverage | Deepak Singh, AWS

Clips
News
More from Cloud AWS re:Invent Coverage

Deepak Singh

VP, Next Gen Developer Experience

AWS

Revolutionizing software development: Deepak Singh on gen AI, AWS Q Developer and the future of coding

Generative AI isn’t just a technological leap — it’s a transformative force reshaping the way software is developed and deployed. At the forefront of this revolution is Deepak Singh, vice president of the Next Generation Developer Experience at Amazon Web Services Inc. In an exclusive conversation at AWS headquarters ahead of its re:Invent conference next week in Las Vegas, Deepak shared his insights into how generative AI and tools such as AWS Q Developer are empowering developers, streamlining workflows and accelerating innovation across industries.“Using gen AI is going to change the way all of us build software,” Deepak remarked

play_circle_outline AWS's re:Invent headquarters and the excitement around GenAI

play_circle_outline AI changing the way software is built, solving customer pain points

play_circle_outline Revolutionizing Coding with Q Developer: Unlocking Productivity and Success Across Multiple Channels

play_circle_outline Speeding up development process, delivering more capabilities to customers

play_circle_outline Analyzing Amazon's Transformation Effort: Leveraging Java17 for Efficiency and Cost Savings

play_circle_outline Introduction of text editor tool for LLM in Q Developer agent

play_circle_outline Fusion of data sets into coding models

play_circle_outline Revolutionizing Coding: GenAI's Role in Democratization and the Essential Q Developer Onboarding Process

Info
Transcript

Deepak Singh, AWS

Deepak Singh

VP, Next Gen Developer Experience AWS

Deepak Singh, vice president of the Next Generation Developer Experience at AWS, discusses how GenAI is transforming software development, aiming to help developers build software faster, more creatively, and spend less time on tasks they don't enjoy. GenAI helps diagnose and fix issues, autonomously execute code, and unblock developers from mundane tasks to focus on building innovative applications. The strategy and vision include voice-activated code, chat capabilities, and enabling users to get back into coding. GenAI supports various channels, tools like ... Read more

explore Keep Exploring

What are your thoughts on the impact of GenAI on productivity in the industry? add

What were the discussions focused on when developing this product? add

What are the current possibilities and future potential of using a developer assistant like Q Developer to write code and assist developers in tasks such as transformation and specification-driven development? add

What benefits can be gained by utilizing GenAI in software development and how can it increase productivity for developers? add

What team at Amazon did we work with to integrate core transformation capabilities into internal systems, resulting in a successful outcome that included efficiency improvements and cost savings? add

What capabilities were added to the software development agent to improve its functionality, specifically in regard to using tools and building its own IDE? add

What are some methods and approaches for accessing and utilizing company data to enhance foundation models for various applications? add

What is the process for getting involved in coding with Amazon Q Developer for someone who hasn't coded in 20 plus years and wants to start using the platform? add

bolt Powered by CUBE AI

Deepak Singh, AWS

search

>> Hello, welcome to theCUBE. I'm John Furrier, host of theCUBE here in Seattle with AWS's in their headquarters. The building's name is re:Invent. It's ironic because re:Invent's coming up. Deepak Singh is here, vice president of the next Generation Developer Experience at AWS, CUBE alumni. We've talked about photography, containers, Kubernetes, and now with CUBE, really rocking and rolling. It's hot. Good to see you.

Deepak Singh

>> Yeah, likewise. Great.

>> Thanks for coming on theCUBE inside the AWS building, I'm inside the ropes.

Deepak Singh

>> Last time we did this, I think was at the re:Invent show floor, so this is a little different.

>> We'll be at re:Invent this year in the Venetian, the press area. We got a lot of interviews lined up, but it's going to be super exciting because GenAI has really brought productivity to everybody. I just did a podcast with Dave Vellante for our weekly pod and the title is Tech Party Time. Everyone's having a good time. There's energy up and down the stack. I'm just talking to Dave Brown. It feels like the nineties you're in networking and compute. If you're in software, it's a Cambrian explosion of new capabilities. The productivity, just the action is off the charts. I've never seen it in my career.

Deepak Singh

>> Yeah I got into GenAI a couple of years ago, and I said at the time... I think when I took over this role, I put this on LinkedIn, it's the last time I felt this excited about any technology was when EC2 got launched. That was before I joined AWS. And a year and a half after that, I was here. This time, right in the middle of it, right from the beginning. I am convinced, and we are starting to see a lot of evidence of this, is that using GenAI is going to change the way all of us build software. It's going to make us build software faster, get more creative, spend less time doing things that we don't like doing. So a lot of the way we think about a developer experience with GenAI is, how can we use this phenomenal technology to unblock our developers from building software, doing the things that they like to do?

>> And I love your title, Next Generation Developer Experience. That really kind of speaks what it is, and there's a lot in there to discuss. And certainly, I want to... And you have a unique perspective. You would doing all the platform engineering thinking around containers and all that stuff that was going into the microservices. And again, web services is a beautiful setup for GenAI, but I want to set it up with my conversation with Werner Vogels two days ago on the Lambda's 10th anniversary at Exclusive 101. And it was a wide ranging conversation, was a lot of fun. Werner's also very animated. But what was interesting is he said when Lambda came out, when Serverless came out, you didn't know what was going to happen, and there was a before and after moment of that where you were doing stuff with the cloud, but then it was growing and more stuff had to get done, and Serverless just changed the game and made distributed computing happen. But he talked about the PRFAQ, and this is where I want to get your thoughts because it made me realize there was a moment there. When you get a lot of people in the room, they go, "Wow." And I asked them a question. I said, "You guys debate, align, and commit. That's the Amazonian way. So what was the debate like in the room when you guys are putting together the working backwards document for Lambda." And he goes, "No one debated. They all were like, 'Wow, this is going to be great.' They kind of knew. They connected the dots. It was all excitement because they understood what it meant." And then obviously it's historic because Serverless changed the game significantly. What's your GenAI version of that? Because I agree with you, there is a before moment where you now go, you don't got to be a rocket scientist. I see all this greatness coming down the pike. What is that moment for you? Because this is like I can see 10 things. What do you see?

Deepak Singh

>> The discussion when we started this was less about what will make this... Is this going to be magical? What do we need to do? It was very clear that this was transformative, or most of us had conviction about it. The big question we were asking each other was, what is the thing that we can do for customers that makes it useful for them today? What are the things we can do for them that makes this technology, which we know is going to keep evolving at this crazy rate, but can help them move forward today? So in fact, the first thing that the product manager on my team who wrote this up was this idea. There were two ideas that came up, and they're both capabilities and Q Developer today. The first one was, how can we help customers diagnose issues, and then help them understand how they can fix them? So a great example of this is a Q Developer in the AWS console. In fact, people get very hung up on talking about writing code, but actually some of the most interesting capabilities in Q Developer are not about writing code. It's about understanding why your application may not be working. So you end the Lambda console and you get this error talking about Lambda. It works really well with Lambda, by the way. And using Q Developer, you can figure out... And Q Developer will tell you what that error is in human language so you don't have to go figure out from documentation, or go somewhere online and start Googling. It'll tell you right there what it is. It'll also tell you how you can fix it. And that, it sounds very simple, but you do that 20 times a day and suddenly you're moving that much faster. So it's small things like that. The other thing that was part of the original, as we had these debates like we do, that became sort of the yes, we need to do this, is what over time became the Q Agent for software development, which is can you give an AI system a problem? And the problem could be an issue in a backlog, which is I want to add a new API to this application. And what Q Developer does is it goes ahead and looks at the rest of your application, sees what's there, how you've written it. Comes back to you with a plan on what code needs to be written, what code needs to be deleted, what new files need to be written. And if you agree, it goes ahead and executes that autonomously on your behalf, creates desks and everything, right? And those were among the first ideas that we discussed. There were many, but these were the two as we started having these debates and discussing it was less about, oh, we have to do this. That was obvious. What do we have to do? And these were the kinds of things that came out, and it all boiled down to, what are the things which we can do that can help developers move faster, help them get unblocked?

>> So those are two meaty examples. There's a lot of meat on the bones. It's not just sizzling. There's a lot of steak on those problems. Those are hard problems.

Deepak Singh

>> They're hard problems. They're important problems. I think they can do a lot of things that look very glitzy, but very often, sometimes the most useful things, in this case, some of them, I guess it sounds glitzy as well, are things that they just help our developers.

>> Yeah. Well, I love the vision. Let's step back for a second. Give me the strategy and vision for how you guys see the transformation and software development happening. Because agents are here. You mentioned you have one. That's only going to be more sub-agents, multiple agents. You're going to have all kinds of... I've been hearing conversations around LLM routing, model routing. It makes sense because you want to integrate and have a runtime kind of a... I hate to use the word compiler, but things kind of work together. You got to put code together. What is the vision?

Deepak Singh

>> Yeah, the word I like to think about and the way I like to think about it is, what are the things we can do for our customers taking into account everything that we have available to us today? And that's not just the LLMs. It's other things that we do. In many ways, making LLMs useful for developers is understanding what customer pain points exist, understanding what AWS can uniquely do for them that they can't necessarily just do themselves. There's a lot more to making LLMs useful for building applications than just the LLM, while acknowledging that every few months the underlying technology, the underlying foundation models are going to get that much better. And we have taken full advantage of that with bigger context windows, better reasoning capabilities. But I think that being grounded on the customer problem remains the key. I'll give you some examples. So we started simple, which was, hey.... And I'm going to talk about code here, is right now, what we can help customers with is helping them with auto-completion of code. That's how most coding assistants started. But code is only a small part of what developers do. So we switched our mindset to no, we need to build developer assistants. So that evolved from just help me type faster to, can I have a chat window on the side where I can ask more complex questions, where it can help me with even have a debate with, is this the right way to think about this problem? Allowing you to highlight a piece of code and asking, what does this do? Right? And then continue to evolve that, which includes building these agents that are goal-seeking where you give it a task and it goes ahead and does it autonomously. And today, we are bringing all of this together. So as recently as a month ago with Q Developer, you either did autocomplete or you went to this chat window, but people want both capabilities within flow because they like what they get in the chat window and they like the flow of being able to type. So we added this thing called inline chat where you start typing but you want to send a bigger prompt and a more complex query, taking into account all your files and all the ones that are open. And you hit Command-I on a Mac, and voila, you can send much more complex prompts through. And this is just one example of our customers-

>> Because you're at the point of a coding point of attack, I would say. A sports analogy, but you're in the point of coding-

Deepak Singh

>> Yes....

>> the context switching break or disruption.

Deepak Singh

>> It slows down your workflow. So our customers told us they like the workflow nature, because that's how they write code today, of inline coding, but they like the richness of the chat context, so we brought them together with inline chat. And it feels small, but it's actually quite transformational for their work.

>> What other partners, or what other systems or partner systems or data connections do you see? Because we've been hearing a lot from Matt, Swami, and others on your teammates there saying meet them where they are, right? They're in Slack. Okay, do some stuff in Slack if they're over here, if I'm using Connect, I'm in a call center. The customers are doing everything in multi-modes, pun intended. Not just data modes, but they're context switching multiple tools. They don't wake up saying I'm in this one thing, and it's this now multi-purpose environments.

Deepak Singh

>> So we support many channels from an AWS context if the channels our customers use the most, so your IDE, which could be any IDE, it could be VS Code, it could be IntelliJ, it could be a Jupyter Notebook. Where you are, we want you to use our plugins. You can use Q Developer inside Slack, for example. So you can go into Slack and say @AWS, answer this question. On your own shell... People like me still love using a shell in the CLI. You can ask questions there. You can go into the AWS marketing website or into a documentation page and ask questions. So from an AWS customer perspective, we asked us where do they sit, and we will bring Q to you. But yesterday, you may have seen the announcement. Wiz and Datadog both announced the general availability, I think, of their plugins for Amazon Q, which means you're in the AWS console and you're doing something, and you have a security issue and you're a Wiz customer, you can invoke the Wiz extension and send that query over to Wiz where Wiz will analyze a security issue and get back to you back in the AWS console. So we are putting Q in the channels that our customers are using.

>> You're creating a trust connection between delegation of task versus some black box.

Deepak Singh

>> Correct.

>> That's a changes the game because now you have end-to-end visibility.

Deepak Singh

>> And over time, talking if you strategy, you'll see many more capabilities like that from other partners coming into Q. There are other channels that we are exploring with the idea of, where do we meet the customers where they are?

>> Okay, so if you can look out and imagine a steady state of success with Q, it doesn't have to be long term, it could be medium, short, medium-term, near future, what does it look like? It's just kind of like voice-activated code? Is it more of in line with the IDE? Is it enabling people to get back at the coding? How does a guy like me who hasn't coded in 20 years get back in the game? I feel like I'm rusty. Does Q help me get going? Is it going to be business people, business as code versus infrastructure as code? What's your vision there?

Deepak Singh

>> So I'll start off with what is possible today because it draws those lines. Today, you have customers like the BT Group that are using Q Developer to write code. They're seeing 37% acceptance rates. There are people like the National Australia Bank that are seeing 50% acceptance rates. What that tells me is that for the developers, a significant chunk of their code is now being written using a developer assistant. You have heard the example of Amazon where we did this massive transformation of a Java code where we went from an older version of Java to Java17, saving 4,500 developers years worth of work and $260 million run rate of cost savings. This is possible right now. The latter one is enabled by agents. Where this is going to go is that a lot of the blocking tasks like transformation of a version are going to be handled for you by a developer assistant, by an AI system like Q Developer. And Q Developer will help your developers get free time to go be creative, to go build these applications. But even there, I recently wrote an article about... Effectively what we were talking about is what I call specification-driven development, which is you are giving the AI system a goal. It is only the developer who understands the system what needs to be built. So they need to have systems thinking, talking on microservices and complex systems, but they don't need to know every single API call. They don't need to know every single AWS service in detail. The AI system can help them implement it in Dynamo even if they've never used Dynamo, but they know they need that kind of-

>> So syntax help, that big-time help. So they know architecture, they can execute.

Deepak Singh

>> They can think about the system they're building, the kind, and they can give requirements to one of these agents, like the Q Agent for software development. And the agent will then look at the specifications, collaborate with them, and then execute it autonomously. And that's happening. It's going to happen really quickly. And what we find are our customers that are increasingly embracing that view of the world.

>> I want to get into that because I want to get into also the numbers, which I thought was an error, but it's in years, not in days. The systems thinking is huge because if you think about what you just said conceptually, to do a systems holistic architecture or solution plan and then build an architecture around it could take literally a half a year, or just a lot of time. And then you got to plan it out, and then start coding it, test it. Just the blocking and tackling, chopping wood, carrying the water to make that happen, it's just off the charts. And now, you can just put it together, move faster. Scope the scale of the speed there because I think this is... System thinking, I think will be this wave's iteration, or design thinking. Remember the one that was hot, these hot trends? We still design thinking it's relevant to UX, of course, but this is a systems mindset. understand consequences and how things work.

Deepak Singh

>> And you actually can think about it at an organizational level. Even simple examples like if you are... Today, developers spend only some percentage of time inside the IDE because they're spending a lot of time doing other things like reading documentation, trying to debug an issue, or doing a transformation. If all of that got speeded up significantly, like you just heard with the Java example, that gives them more time to be creative inside the IDE, which is I'm using that as a tool for development. And if you can speed up how much they can accomplish with that tool for development with things like the agent for software development, you can now build more. Very often you'll hear customers say, if you're a company, your customers want you to deliver more features. If you are a bank, your customers want more products or faster banking. The ability to do that, deliver more capabilities to your customers, deliver them at the higher quality, that's what GenAI is going to help you do as builders, and we're excited about that.

>> I think about the Q has been getting a lot of great reviews, so congratulations. There's a couple of notable things here I want to just get your thoughts on. One was the redesign for the developer agent. You mentioned reasoning earlier. I want to talk about that. But first, you guys have been using it internally, which is the classic Amazonian move. You identify your own problems, and you're at scale. What was... 4,500 years of manual work, and 260 million annually. I said this morning, did you get that right? No, it's years. Obviously it's man years and you kind of do the calculations, but it's a big number. I'm not trying to interrogate the number. It's more of the size and scope and the order of magnitude. It's like, oh my god, that's like... I was thinking maybe days, weeks.

Deepak Singh

>> It's also a function of the scale of that.

>> It's just huge.

Deepak Singh

>>

>> Take us through how that happened, and then I want to jump into the redesign, some of the speed stuff you guys have been doing.

Deepak Singh

>> So one the things that made it happen, and this goes into the systems thinking, it's not just some magic thing that you switch there, you press a button and pixie dust happens. We had a really good partnership, and we've... And this is what we're doing with our customers as well. It was a great learning for us, and we have applied that to our customers. Where we worked with... At Amazon, there's a steam called the Amazon Software Builder Experience team. Their job and mission is to enable our software builders inside the company. We worked with them to take the core transformation capabilities that we are developed and integrate them into the systems the way things work inside Amazon. And so they were able to, from an end user perspective, make it very, very simple and very easy to do these transformations using... So they took what we had, they built around it, integrated into our internal systems, and that's what gave you that level of success. Some of it is due to the scale of our internal code base . And Java17 is that much more efficient than Java8, which is where the dollar savings come from. So yeah, and we were able to measure it because we happened to have those

>> And some people call it hygiene, some people call it cleanliness, data cleaning, code cleaning, but essentially you went into the code base and said, "Okay, where our we version, and let's make sure we're up to speed." You can do that anywhere. You can point Q at security, configurations. This is where this is killer.

Deepak Singh

>> Yeah, we've started seeing customers. There's this insurance company that had spring boot application and start upgrading them, and they said they're doing it, I forget the exact number, I think about 36% faster than any other transformation effort that they ever had.

>> Yeah. Dave and I were just joking on our podcast. I said, "Dave, this whole legacy modernization is happening." And I go, "You know that's code word for putting a wrapper around it to use old, old-school language?" But just migrating, and then making things run, it used to be compatibility mode means you have to right gateways to kind of make software work, but now you can actually integrate old code bases. So you got to kind refresh them. So you can actually modernize legacy stuff, and then integrate it into the modern. You don't have to kill the old to bring in the new.

Deepak Singh

>> Yeah. And you can upgrade it more easily also. If you have code sitting in one language and most of your developers on another area, you can start doing that. And AI systems are pretty good at doing that and actually telling you what the old code did, the explainability part of it. It's one of the most commonly used capabilities we have, is tell me what this code does.

>> Yeah, that's a great feature. In fact, one of the use cases I hear from a lot of the customers that you guys have, and prospects and user customers, is we're using AI to figure out what our AI strategy should be. Because a lot of the times, they never had the opportunity to actually do run use case scenarios around planning. So it's like, oh my god, I all this data. And first of all, what do we have? And then they get in, and then they see new products that come out of the code and/or the data. So it's interesting, a valid call out there. So I want to ask you something that's more product oriented. So I want to get into the reasoning, but first I want to ask you, how has this developer progression, this change over transformation changed the role of the product manager? Because I can see the product manager's game speeding up big time because now with the ability to kind of look at code, there's a lot more velocity going on in the product management side. So the product manager who has to do the normal, I won't say slow, but methodical, this is our product requirements, PRDs, and the old school... You have to kind of get faster on that. How is that changing? Because I can almost see the development going faster than the product specs.

Deepak Singh

>> As a former product manager,-

>> That's why I asked.

Deepak Singh

>> We started this conversation by talking about PRFAQs. I have this saying in my org. I may have heard this from Matt Garman back in the day, which is you should always have more PRFAQs than you can actually write software for. I think it's very interesting. Now you're reaching, you can maybe catch up with that, which means you'd write more PRFAQs and write them faster. Actually, our product managers are excited. They're as excited as they can be because they have a lot of ideas. And sometimes some of those ideas can be prototyped without writing code. I don't know if you talked to anybody about Q Business recently, but within Q Business we have this idea, this thing called Q Applications, Q apps. We also have PartyRock. Those are great prototyping tools. They're great for, hey, I have an idea. How do I convert this into a potential design of something like that? Which is helping them iterate through their own ideas, taking customer feedback and being able to distill it down, which then helps them write better PR facts for our engineering organization.

>> Do you have a automated PRFAQ agent yet?

Deepak Singh

>> I am... Shh. I'm almost certain that somebody has taken Q Apps and written one of those. I... Yeah.

>> I know for a fact that PartyRock and the tools you're mentioning have been used by Amazonians on the business side. I've heard people say, "Hey, I now know how to interface into the SQL databases. I just tell them what I want as a result and it tells me what the SQL code is." And so you're seeing harmony between teams because the friction's coming away. And so on the product management side, I think there's always backlog of things to work on. To your point about Matt's comment about they should have everything stuck in there and prioritization. The bottleneck was coding and all the process. So now if you take that friction away and they could more time coding, then the product managers deal with velocity. It's like whitewater rafting. You're going faster.

Deepak Singh

>> Yeah, product managers are itching to go even faster. We have all these tools now. And as all of this comes together in the next few years, I see this as the creative juices of all the builders out there getting unleashed.

>> And the job satisfaction would be higher. Think about just things are moving off the plate faster. Okay, great. So that's a good... We're going to dig into that at re:Invent heavily. The whole productivity boom that's coming, I think it's going to be revolutionary. I think it's going to be something we've never seen before. New apps going to come, and just these tools are just the beginning. All right, so the redesign behind the Q Developer agent, this has got reasoning. And take us through that. I saw the blog post. Was it a month ago or so?

Deepak Singh

>> A couple of months ago now. Time flies.

>> Feels like yesterday. Okay, so take us through the redesign because this is notable.

Deepak Singh

>> Yeah, so some of it is just LLMs getting better and allowing more things. Some of it is understanding what our customer's expectations are. Some of it is our own evolution of thinking around what an agent is. So one of the things that became very clear to us as the people started using the software development agent and using it in the wild, as I like to say, in anger, was that you were... LLMs and foundation models in general were getting bigger context windows. They were getting more capabilities, and one of those interesting capabilities was the ability to use tools. So one of the things... There's a bunch of things we did with that agent, but one specific capability we added to the software development agent was building its own IDE. This is like a text editor. We call it text code. And what it allows the agent to do is the LLM can now open files, close files, almost like highlight text, delete text, add text. It's all doing that behind the scenes. It's the LLM that's doing it, not a human. But the LLM now has access to a text editor, which allows it to do these tasks and almost C code interact with it. And this whole idea of making tools available to an LLM as part of the agent design... And this is one very specific example of something that we did, amongst other capabilities that made that agent get so much better. So for example, at the time that we wrote this blog post and published it from the previous version on SWB bench verified, we were 51% better. On SWB full, we were 43% better. And we've continued to evolve the agent. We actually have newer versions now, and I feel like, again, it's a fly deal.

>> It's getting more accurate as the price performance of token costs and price forms of value is-

Deepak Singh

>> It's a combination. It's capabilities of the underlying foundation models. It's capabilities being built in that allow some of these tool actions to take place. It is understanding what the system's design should be, but it's all being driven by the things that we see our customers doing. How are they interacting with this? What are the problems they're trying to solve that help us go back and say these are the, help our scientists go think about, okay, these are the experiments we need to do. For example, let's give our LLM an access to our text editor, as an example.

>> Talk about the fusion of data sets into the coding model. Because I've always been saying for years on theCUBE, you got horizontal scale clouds, great, but vertical specialization with the data is... We did a blog post this last week. We said Jamie Dimon is going to be Sam Altman's next competitor. Just inferring with the sensational headline was is that JP Morgan and JP Morgan Chase, they have massive amounts of data and that's their value, so they're going to probably end up coding against it. So there's this data as code, you got business as code. And DevOps showed us a lot, showed us at that platform level that infrastructure code is a good thing. APIs were great. Now you got GenAI connecting in on top of the API. So you've got this whole new wave coming in where people are going to want to jump in and start coding. So you're going to start to see smaller language models integrating with the large language models, foundational models, computer vision. Computer vision is going to be massive. So you have all kinds of multimodal AI, but the little model's got to be had the proprietary IP behind it, intellectual property.

Deepak Singh

>> The way to talk about it is actually the data that a company has. How it access it, how you make it available to your foundation models, whether it's through distillation, whether it's through fine-tuning, whether it's through RAG, all of them are methods and approaches, or even training your own model. I think you will see companies use all of these approaches where they will start with knowledge bases in Bedrock, for example, which are data that only you have access to that's going to make your foundation model more effective in whatever application you're providing to your customers. We see that. For example, with Q Developer, we allow you to customize your responses based on your code base and your... So it's not a generalized response or generalized suggestion. It's a suggestion that's driven by the design of your application and the packages that you use. People might do fine-tuning. They might take an existing model, like we allow for something like Haiku, or for some of the other models that we have in Bedrock, you can fine-tune them with your own data and make them more relevant to your customer base, or you could go into SageMaker and train your own model. So I think you'll see the spectrum depending on the type of data you have, that type of application you're building, how well represented that type of problem is in the foundation models. And no, for example, in the healthcare space in particular, you'll actually see lots of custom models because healthcare data is almost completely intellectual property for these pharmaceutical companies.

>> And they get workflows that are well understood, but yet complicated, and bringing technology to that is just going to be much... Takes all the friction out. What was once a hassle siloed systems just now integrating-

Deepak Singh

>> The barrier to entry is going to keep going down, and the ability to then deploy something like that to SageMaker or to Bedrock and in a serverless way is going to be very important.

>> Yeah, you bring up a good point that I think is bigger picture that I think is going to be the re:Invent tone, is impact, whether it's societal impact, better healthcare, or climate change, or just making people more productive. One engineer that I talked to said, "Hey, I love all these new tools. I can code more. I want to code." I said, "What's been the big value?" He goes, "Productivity." He didn't say productivity. He said more beer time. And what he meant by more beer time was, I can drink more beer with my friends. And what he was saying is, "I get more leisure time," which was valuable to him. Some people want to be creative. Maybe they drink a few draft beers and go, "Hey, let's work on this project." So to your point about systems thinking, all that drudgery is gone, or not gone, or reduced, or almost eliminated.

Deepak Singh

>> I've been at AWS for over 16 years and one of the most fun parts about being here is the impact we've been able to have, whether it was when EC2 first came out or when I was working on containers, and now GenAI is. I sit down with our customers and they have all these ideas on the things they can do, which they either had before but didn't have a mechanism or a way to get them done, or they just didn't have the time, and that suddenly they have tools and things that are unblocking them. And if you could even do half of... If they can accomplish half of what they want to do

>> I remember when people are indifferent, they acquiesce. And if it's too much, as they say, if the juice isn't worth the squeeze, they just abandon it or just eh

Deepak Singh

>> Actually, I'll give you a very good example. This is a predecessor to PartyRock and Q Apps that we had that was built internally, and the learnings from that eventually ended up in those public products. But we had a bunch of data center technicians who built an app which would probably never have been built because nobody's going to have a software team build one app and maintain it for a small team of data center technicians. But what these folks did was they took all the documentation of their HVAC systems, they operated the HVAC systems in the data center, uploaded them to this AI system, and they basically, with a couple of prompts, built this app that allowed them to enter error codes. The app would then troubleshoot their code, and then provide them the solution based on the documentation. And they're not developers.

>> It went nuts, yeah.

Deepak Singh

>> It's one of those things where something that may never have been done got done because of the fact that there was this GenAI-based system all built on Bedrock.

>> Well, that's the democratization. So let me ask final question. That's why I said a guy like me hasn't coded in 20 plus years and to get back into coding. How do people get involved? Because you have a huge customer base and loyal customer base that knows the console, and some people just... They've been provisioning resources in the cloud, but now you're starting to get a new kind of developer that wants to jump in. So it might not be obvious to folks who haven't used Q. How do they start? Is it the console? Is there an onboarding mechanism? How would you get onto Q?

Deepak Singh

>> It's very easy. One way to do it is go to the aws.amazon.com, Amazon Q Developer detail page, and you can either download our plugins for Visual Studio or for IntelliJ, whichever your favorite IDE is. And you don't even need an Amazon account. You can use something called the Builder ID, which you can sign up for. It's like a social login, and it's free. And you can use Q Developer up to a point. You won't be-

>> The freemium version.

Deepak Singh

>> We limit the number of transformations you can do, et cetera, but you get all those capabilities. Or if you have an AWS account and you log into the console, just log into the console and Qs available to you on the side. And if you buy Q Developer Pro licenses, you can do all kinds of things with it. And even there, we have a very liberal free tier, so you can-

>> kick the tires, getting their hands dirty.

Deepak Singh

>> More than kick the tires. You can do meaningful work. And it's only when... You have a per user license. When you have... Once you sign up a bunch inside an enterprise,-

>> The goal is get immersed into the tool.

Deepak Singh

>> But to get going, you can all just start with the Builder ID. Go home, download it to your laptop, learn how to use these. I feel like any builder, whether they're a developer, an architect, a sysadmin, somebody who's just interested in building should go download Q and learn what you can do with it.

>> Deepak, great to see you. The next generation of environments here, it's going to be reduce the friction, reduce the steps it takes to do the mundane work or undifferentiated heavy lifting, get back to coding with co-assists, co-workers, teammates, whatever you want to call it, and that's the future.

Deepak Singh

>> Yeah. Yeah. And I'm very excited to see what our customers do as we keep evolving.

>> What's going on at re:Invent? What's the big plans?

Deepak Singh

>> 10 more days.

>> You're tight-lipped. The PR team does a good job. I can't get anything out of you guys.

Deepak Singh

>> I think you'll see a lot more about the kinds of things that we've talked about.

>> Great stuff. Thanks for coming up.

Deepak Singh

>> Always great to be here.

>> Good to see you here in the home turf for AWS. It's a road game for theCUBE. We are here at the Amazon headquarters, Amazon Web Services headquarters. The building's called re:Invent. The big conference is coming up. We'll be there with theCUBE. I'm John Furrier, your host of theCUBE. Thanks for watching.