Subscribe
Share
Share
Embed
Podcast on enterprise technologies including cloud native, GenAI, Security, Data Protection and more.
Hi. This is your host Swapnil Bhartiya, and we are here at KubeCon and Cloud Aditcon in Salt Lake City, Utah. And today, we have with us once again Arun Gupta, VP of developer program at Intel. Arun, it's great to have you back on the show.
Arun Gupta:I'm very happy to be here. Thank you.
Swapnil Bhartiya:It's my pleasure. I'm I'm kind of used to this look every November. So talk about no, why are you doing this?
Arun Gupta:Yeah. This I've done this for over a decade now, and this is about no shave November. And that's the movement started many years ago to really promote men's health. You know, so whole idea here here is social, mental, and physical health matters for everybody, but let's use this as a statement to talk about men's health. And really, you know, make sure that you're not worrying alone.
Arun Gupta:And men have a lot of worries in their head. Talk to your friends, talk to your colleagues, talk to your brother, father, whatever, partner. Right? Do a regular workout, eat healthy, have a good social life because that is the way you sustain longer.
Swapnil Bhartiya:Since you have opened a can of worms, so I will not just jump from this topic to other topic. I'm also a man. We are unlike women, we don't like to talk about our problems. Sometimes we are not raised in that way. Sometimes even if, and I'm from India, you know, and there's a joke also that in our time, nobody had depression, you know, because a slap of her dad will kick all the depression out of, you know.
Swapnil Bhartiya:And then if you do say I'm depressed, you're like, just, you know, forget about that. Just move on. Right. But now it is becoming an important now let's just forget about a lot of social thing. Let's just talk about workspace.
Swapnil Bhartiya:I used to have a very good friend at Intel and he was and his philosophy was also to unplug yourself. Even sometime the work related stress can be stressed. So while, you know, what is your philosophy? Because you, I am aware of you talk about meditation a lot, you talk about yoga a lot. So what is your philosophy to maintain a balance between work life and personal life?
Swapnil Bhartiya:Because that also, at times, contribute to mental health.
Arun Gupta:Very well so. I mean, I think you're absolutely right. It's very important to have that work life harmony. People call it as balance. I call it as a harmony.
Arun Gupta:You know, because when you are working, then you are all in. But then when you are not working, then you should be all out. And, frankly, like, I like to run every day. I do lifting, you know, twice a week, 4 to 5 times running every day in the morning. And I've changed my meeting schedules, team meetings, customer meetings that no.
Arun Gupta:Self care is important. I gotta take care of it. So that's super important to me. In day to day meetings, you know, we do moments like moments of gratitude. Because oftentimes, you start a meeting with, that is boring.
Arun Gupta:That is not working. But having those simple moments of gratitude allows people to connect better. We take a very mindfulness approach. And in that, what we say is, okay. Fine.
Arun Gupta:We need to tackle this problem. What is in our impact and control, and what can we do? Because I cannot control the other party. There's a Hindi song that I've been listening for the last few days. It's been ramming in my head.
Arun Gupta:It goes like this. That to me is mindfulness.
Swapnil Bhartiya:Can I interrupt you for a second? For a lot of listeners who don't understand Hindi words, because this is a very deep, you know, song that if you can just quickly translate into English as well I can. What it means so that Yeah. They can also understand the depth of what you're saying. Yeah.
Arun Gupta:So what this song really means is you don't know what is ahead of you. You don't know what has you know, you don't have control over what is gone behind you. What you have is the current moment. And I think applying that philosophy at your work, in your personal life goes a long, long way. Because if you are one foot is in the present sorry.
Arun Gupta:One foot is in the past, one foot is in the future, then you're pissing on the present. Let's make the current moment meaningful and do your best.
Swapnil Bhartiya:Laul, let's also talk about OPR, what it is, and, also, if we can talk about briefly before OPR, how important is open source for Intel?
Arun Gupta:Yeah. I mean, Intel has always done open source. You know, I was looking at it. We have been doing open source for, gosh, 40 plus years to first contribution to GRU Compiler. And the open source is very much crucial and strategic to our entire foundational approach.
Arun Gupta:We believe, open source, you know, allows multiple players to contribute and collaborate together. It creates a fair, equitable playground. Innovation happens faster, you know, in an open world as opposed to a walled garden. I was in Serengeti earlier this year. And there, it was amazing to see a lion and a gazelle and a giraffe and a zebra and a hippo and a rhino and a ox all in a open world.
Arun Gupta:No borders, no boundaries. And Serengeti tagline is Serengeti shall never die. And that's exactly what we see open source as. Open source shall never die. I can never see a line in a cage now.
Arun Gupta:Very much like, you know, I don't see myself in a walled garden. So in that sense, that entire has been a philosophy for us. We contribute to 300 plus open source projects. You know, we are part of 700 plus open source foundations and standard bodies. You know, we do this chop wood carry water work.
Arun Gupta:It's not just the code contributions, sponsoring events, you know, being at the governing board position, reviewing pull requests, all sorts of fun stuff. So very much foundation of what we do.
Swapnil Bhartiya:When we look at open source, you know, of course, we look at folks who are building stuff and also folks who are consuming, you know, their direct, you know, the developers, you know, maintainers, you know, creators. How does Intel look at developers? Because for a long time, like a couple of years, we start talking, develop it, start our DevOps and etcetera, all those other labels and personas, but not but now we are hearing a lot about developer experience a lot about that. So talk about Intel's own journey to engage with developers, and how do these project further help that cause?
Arun Gupta:Yeah. I mean, developers has always been very key, to Intel. You know, end of the day, the reason we contribute to these 300 plus open source projects is our customers, our developers, when they go build an application on a hyperscaler or when they buy a laptop from Best Buy. You know, end of the day, they download these open source projects, and we wanna make sure that they are provide the most optimal experience on that VM or a laptop out of the box. So that developer obsession has always been there.
Arun Gupta:Now with my recent charter of developer programs, you know, I have full visibility into what's happening a little bit more deeply. We engage with, you know, hundreds of universities. You know, we do hackathons. We do workshops. We write code samples.
Arun Gupta:We create teaching kits that we give to professors who are then teaching Intel software technologies as a full one semester course. So that when these developers, young developers, they come out in the industry, they're ready to kinda mobilize it. And that's the reason we are here, to talk to all the developers, you know, all the builders on what kind of fun stuff you wanna build, listening to their feedback, and integrating that back into our project, into our products is fundamentally important to Intel.
Swapnil Bhartiya:Now I also I want to just go a bit deeper into OPI project as well for those who don't know or those who do know. Talk about what it is and what is the status today?
Arun Gupta:So, yeah, OPI is open platform for enterprise AI. It's very much part of the enterprise AI strategy for Intel. And, essentially, we created this project, you know, about 8 months ago when we realized we are running multiple Gen AI efforts across the company, and we started talking to our partners, customers. They were running similar efforts. So OPA in a high level is a open platform that allows you to simplify Gen AI app development.
Arun Gupta:That's what it is. And it's fully aligned with cloud native principles. At a very high level, OPA has 2 main core components. 1 is the Gen AI comps. That's the GitHub repo.
Arun Gupta:And that has a bunch of component level microservices, embedding, retriever, re ranker, fine tuning, LLM, vector database. These are all the components you need when you're building a Gen AI application. Leveraging these microservices, we have blueprints, which is Gen AI examples. So like a RAD chatbot or a audio q and a or a code translator, or a code generator, or a doc summarizer. So we have about 20 plus such code samples, which are blueprints that customers can take and deploy into their application today.
Arun Gupta:Now because this is all fully cloud native, you can run this using Docker Compose on any instance, ec2 or Azure Compute or GCE or GCP, or you can run this as a Helm chart on any of the stock Kubernetes cluster. So we have been working with several hyperscalers very closely for a deeper integration so that go back to developer obsession. Right? Because developer, what they're looking is, how can I create a simple RAB chatbot? We are working with hyperscalers to provide OPR solutions in their marketplace, where they say, I want a Gen AI chatbot.
Arun Gupta:Click, and we're gonna have this entirely deployed in your environment. Say, in case of AWS, it'll be deployed on EKS, Amazon EKS, which is their managed service. So in that sense, starting from what is your concept of what Gen AI application you wanna build, do this in a cloud native manner, keep it hyperscaler agnostic, keep it vendor agnostic. Even though Intel created project, we have seen very good traction. We have 45 plus partners to OPMO, including AMD, who's gonna do the testing and the validation.
Arun Gupta:We're talking to other hardware vendors as well. So we've seen pretty good traction over the course.
Swapnil Bhartiya:What kind of evolution you're seeing of KubeCon as the market is also evolving? Because early days, Kubernetes was for it it stayed less workloads, not. So our stateful workloads. And just like the external, the original idea of Kubernetes or Borg was different. And now, suddenly, when you put something in open source, the way people start using it, you cannot even imagine.
Swapnil Bhartiya:And same thing is happening. At the edge, they are like even small Kubernetes. So how do you have seen the whole evolution of Kubernetes and how you see Intel continues to play a very key role in this space?
Arun Gupta:Yeah. I mean, Kubernetes has definitely continued to evolve. You know, as you said, right right rightly yourself. Right? State less, state full, and then over a period of time, you know, now supporting all the AI workloads.
Arun Gupta:There was a full blown cloud native AI day. You know, on Monday as a colo day. So lots of interesting and engaging talks over there. And a lot of the time when we are talking to customers or developers and you say, hey, where are you deploying these AI applications? In the cloud, I said, but how in the cloud?
Arun Gupta:Kubernetes. So the it's a given de facto compute platform where customers are deploying their AI applications because everything that Kubernetes offers as a matter of fact, you know, cluster management, state management, you know, auto healing capabilities, and how do you place the workloads correctly. All of that is good stuff that you would expect. You know, you don't wanna build your own orchestrator that is cloud agnostic. Kubernetes already offers that.
Arun Gupta:So on top of Highscaler, it gives you that compute layer, a universal compute layer. You just deploy your workload and then have your data scientist. So what I'm seeing on the floor and in the talks is how you are bridging the gap from MLOps or KubernetesOps, where your data scientists are creating the model, fine tuning the model. They're doing these experiments and they say, okay, this works good. This gives us the right result that we need, good predictability, good accuracy.
Arun Gupta:And now take this into production. And the platform engineering team that is sitting on top of Kubernetes is really taking that experiment and saying, naham, we're just gonna productionize it because everything is containers, so it just runs and scales over there. For Intel, it's super critical. You know, we are very customer obsessed. So one of the things that we worked on, you know, we have been part of the Kubernetes community from the very beginning itself.
Arun Gupta:And that's the reason Intel allows me to be their representative on the governing board. And they're very supportive for me to be the governing board chair as well. Because when I became the governing board chair, the idea was that, oh, something has to be given away in order for me to take up this new role. So Intel is accommodating in that sense. They gave me more people to run those kind of a things.
Arun Gupta:And then from the technology perspective, you know, we continue to be among the top 10 contributors to Kubernetes. So, over the last couple of years, we have worked on this thing called as DRA or device resource allocation. And which is sort of a vendor agnostic way on how you can leverage, you know, when Kubernetes has to do the allocation, or place their workloads on a certain device. We work with Google and Nvidia on this. On and that's the beauty of Kubernetes.
Arun Gupta:Right? It brings all those competitors together. And now we are co competing. So that's the beauty of it. That, you know so that's where we continue to push, this API of DRA was announced as alpha last year, KubeCon, and now it's beta.
Arun Gupta:And going forward, once it gets to the GA phase, then, of course, lot more wider adoption.
Swapnil Bhartiya:No. That's the beauty. And, the open source is that, your competitors developers versus the code bank that you have created, you know. And and you don't do hire everybody because you get the best of the breed folks to work on your code base. Yes.
Swapnil Bhartiya:Your sales teams will compete in the market, but developers, they just look at that one code base and then gauge. Now can you also talk a bit about since we are talking OPI and AI or JEN AI in general, I don't wanna name any company, but, you know, open source and Gen AI is not as easy as the LAMP stack that we used to talk about. So talk a bit about what does open source really mean in the AI Gen AI space? What are the challenges in kind of trying to find a utopia of open source in the AI space?
Arun Gupta:I think you said utopia. Right? That's the right word. When will we get to utopia? We don't know that yet.
Arun Gupta:And this is definitely in early phases of Gen AI. Like, if you think think about it, it's barely 2 years that ChatGPT was launched. And, of course, that explosion has happened since then. And in terms of open source and Gen AI, really what they're looking at it is, is your model open source? Is your weight open source?
Arun Gupta:Is your data open source? Is your vector data vector database open source? And is it a open source? Is it open API? What is the access over there?
Arun Gupta:So Linux Foundation, there's a GenAI Commons working group in LFAI and data. They have created this thing called as model openness framework. So model is the biggest part of it. So they're saying, okay, take a look at your model and rate it on the MOF framework. And that kinda gives you how open your framework is.
Arun Gupta:So that's a good way to judge it. But, you know, if I bring it back to OPA for a second, OPA is a open platform. So as much as you are doing integration with, you know, Milverse or Quadrant or Chroma, we're also doing integration with Pinecone. You know, as much as we care about integration with any of the back end LLMs like Llama or Mistral, we're doing integration with Cloud as well. And that will come through integration with other managed services.
Arun Gupta:So I think the open versus closed debate is nothing new. We have seen this with Unix, versus, Windows. We have seen this with iPhone versus Android. So this is a very healthy debate, and I think it's a very constructive debate. And I always like to call these as NORA, no one right answer.
Arun Gupta:So there is really no right answer. It depends now what gives you value. There are both downsides and upsides because let's say if your model is totally open. If it's open, then maybe you can explain the results better, but that makes it that much more accessible to the bad actors as well. So how do you prevent that?
Arun Gupta:How do you build guardrails around that? But then if you are on the close side, that means you're really, really reliant upon the vendor to do the innovation for you. So I think there is really no, as we say.
Swapnil Bhartiya:When we look at AI, Gen AI, you know, the adoption is growing. There are some genuine concern which has nothing to do with the core development. It is more about consumption. It is more about ethical AI. It could be more about responsible AI.
Swapnil Bhartiya:Also, sometimes when we look at, you know, chat GPD, with open source, we create the whole creative commons. So you respect my copyright. But some of those things, they don't respect those copyright. So what are your concerns if that is within your scope because you are involved in OPIA that we are you're worried about some of the trends in terms of GNA REI where you feel that, hey. We have we should also address those problems.
Arun Gupta:No. I think it's very important to address those problems. See, because end of the day, if you ask the question to an LLM that show me the picture of a working person, And if that working person always happens to be white, that's an inaccurate representation. Right? How do you explain that model?
Arun Gupta:What is the explainability? You know, what dataset did you train on? So I think Intel, in that sense, have a very strong bias towards an open model, and we really believe in responsible AI and ethical AI. We have deep teams in Intel Labs that work on that element. And how do we make sure this is ethical and responsible?
Arun Gupta:So that's a critical element of it. But and that's where the conflict happens. Right? In case of, closed source vendor, whatever they believe is responsible is responsible. But in case of a open source, you can actually see what responsible means and how they have done this.
Arun Gupta:And also, I think responsibility goes all up and down the stack. And how much of this is available and visible and exposed is a critical element.
Swapnil Bhartiya:So as you said, you know, OPI, you've it's it's 8 months old, you know. It's still a baby. Can you talk about, what kind of folks are looking at it? If you can share, depending on name, use cases where you're seeing OPI is already been used.
Arun Gupta:I think one of the most exciting use case for us is, you know, for customers, it's very easy to start with, you know, a managed service. And, you know, they get a quick response. It's very easy for them to get started. But then they realize, oh, the app is getting popular, but now they're charged cost per token. And that becomes expensive for them right away.
Arun Gupta:And the performance scales, all of that is good part, but they wanna bring the TCO down. And that's where we are seeing a lot of OPR adoption. And what happens is, if you look at a managed service, you know, Azure Open AI, for example, not to pick on them, but just one example. It is, embedding, a retriever, a re ranker, a LLM, and a vector database all combined into 1 with a single API abstraction. So with that, where what we're seeing traction is our sales team is excited because now they can bring that API and they can break it apart into these multiple microservices that exist in Opria, and they can run it on a Xeon instance on Azure.
Arun Gupta:2 big advantages. First of all, because of open source, the cost is significantly down. And second is the customer have full control of their data. And we are working with some really large customers who really, really wanna hold on to their data, and this is a really good solution for them.
Swapnil Bhartiya:Very well, sir. Thank you so much. Now I wanna change the topic a bit, and, I wanna talk about your book. Talk a bit about your book, and that's when you'll go from there.
Arun Gupta:Yeah. I'm very excited about it. So, I over the last year or so, I've been working on this book called as Fostering Open Source Culture. And this book is really based upon my experience, by building that culture now at Intel, but prior to that at Apple, Amazon, Sun, Red Hat, Couchbase, and all of these companies. So part of the book is my experience on what I've seen as the trusted recipes, but the more exciting part of it is 40 plus case studies in the book.
Arun Gupta:And these case studies are by 50 plus contributing authors from all around the world. You will see automotive, financial, media, technology, retailer, all sorts of industries. Because no matter where you are, you are using software. And if you're using software, you are using open source. So it's very important to build that open culture in your company.
Arun Gupta:And this book kinda gives you all of those 40 plus recipes that you can leverage, and more importantly, a framework that I don't know where to start. So it gives a very opinionated approach that I've seen work again and again. And I'm quite excited about it and looking forward to it.
Swapnil Bhartiya:When you talk about open source culture and open source, the the the first thing is that you basically said, you know, you have talked to different, you know, 40 plus in our use case. So that already answered my question because I was going to ask, Gus, how a company consume open source also depends on what kind of business they are in. You cannot have the same recipe for everybody. But are there any common threats, common patterns? So irrespective of what kind of company you are, what size of company you are, these principles will work for you if you are adopting open source.
Arun Gupta:Correct. No. Very true. I know, usually, you start with internal facing tools or infrastructure tools. Like, you don't wanna write a new Kubernetes.
Arun Gupta:Right? Doesn't buy you anything. And if you wanna run something on a cloud agnostic compute platform, Kubernetes is the by default answer. And if you're using Kubernetes, and if you have to scratch your own edge, let's say that you have one problem you wanna solve, 99% of the problem is already solved. You just put a maintainer over there and say we're gonna contribute it upstream.
Arun Gupta:And the whole value premise is you don't wanna fork Kubernetes because then you're missing out on all the innovation, and then there's a technical debt in your company to maintain that internal fork. So it's lot better, lot more cost effective, lot more impactful for you to contribute it out in the upstream and continue to use that. So in that sense, that internal engagement, internal developer tools, and all of that is a great area to start with.
Swapnil Bhartiya:What kind of folks should read this book?
Arun Gupta:This book is actually targeted for anybody and everybody up and down the chain. So I would really my hope is, executives at any company, independent of what industry you are in, this is very targeted to them. It is also targeted at developers on the ground that, hey, we wanna drive this, but we don't know how to get started with this. So it gives you framework, it gives you mindset, it gives you thought process. And independent of how deep or how early you are in your open source journey, everybody will have some nuggets that they will learn.
Arun Gupta:And my hope is people go back to this book again and again, and from those 40 case studies that they read, pick nuggets that they can apply in their daily work. So I think it's a very wide audience, definitely not just targeted at tech audience.
Swapnil Bhartiya:Arun, thank you so much for joining me today. And wide range of topics today from MenHealth all the way to open source to your book. Thanks for great insights as usual, and I look forward talking to you again. Thank you.
Arun Gupta:Thank you so much.