 Hi, this is Hoseppin Bhartiya and today we have with us Arun Gupta, VP and GM of Open Ecosystem at Intel. Arun, it's great to have you on the show. I'm super excited to be here, Swapnil. Yeah, and today we're going to talk about Intel's, you know, donation or contribution of Open FL framework to Linux foundations, AI and data foundation. And of course, we will talk about Intel's open source strategy. I mean, I have been covering Intel and open source from a very early on, even when I was a journalist in India because Intel, though it's known as a hardware company, you folks do a lot of open source either way. So for me, I am not a newcomer to Intel's open source, but a lot of people will be. So I'll talk a bit about Intel open source, of course, Open FL. Before we go there, I want to talk about you also. Talk about your own journey and what led you to join Intel? Well, I've been always like an open source citizen, you know, for over two decades now. My history goes back to Sun Microsystems where we took the company, people process technologies out of the three key elements, you know, brought that cultural change at Sun Microsystems that many years ago, almost 20 years ago. And then I've always stayed on sort of the open source path. That's how the common thread across all my beads. So I went to Red Hat for a couple of years, couch base for a couple of years after that, spent Amazon a few years over there. And then my latest gig was Apple, where I started the open source program office over there. So when Greg reached out that around we want somebody, Greg Lavender over CTO reached out, that around we want somebody to build the open ecosystem team and say, Intel, and what do you do in open source? I'm not fully aware of it. And then when he started telling me the story, I said, wow, that is a beautiful gem. And why is it hiding? So that's the story we wanted to tell. And essentially that's how it came about to be. So if somebody asked me, yeah, the VPGM is sort of the formal title, but really the main title here is the chief storytelling officer. How do we tell the story? What is Intel doing in the open source? And why is it doing it? And how is it doing it? So that's sort of my main element here. And that's how I came about to be just about in a month, I will be about a year at Intel. And I'm super excited, super thrilled. I would say other than maybe Red Hat, this is the only company where everybody in my management chain truly understands the value and the premise of open ecosystem. Thanks for talking a bit about your journey and also highlighting the work Intel is doing. Actually, when you said that, you know, your job is to tell stories. I mean, I am also kind of a journalist, but I'm also an open source advocate. Even if folks know it's very important to repeat the message because sometimes repeating a message is as important. I remember my early days of journalism, it was back in 2005. We are telling people why you should use open source today. Almost everybody is using open source, but a lot of people don't even know companies don't know how to become a good open source citizen. So it's very important to repeat that message. And that's why I'm going to ask the next question, which maybe a lot of understood in the open source community, which is that talking a bit about Intel is technically seen as a hardware company. What does software mean for Intel? And why does it do open source? Well, Intel fosters an open ecosystem strategy to build trust, choice and ensure interoperability across the industry. I mean, yes, Intel is primarily a silicon company, but the way our customers consume that silicon is through software. And as they say, software enabled silicon, software powered silicon enabled. So that's the reason, you know, Intel has been the top corporate contributor to Linux kernel for over 15 years. That's the reason we are one of the top contributors to Kubernetes. That's the reason we are one of the top corporate contributor to open JDK. We are the number three contributor to PyTorch. So those are the projects how our customers consume Intel silicon. So making sure those projects are run in an optimum and efficient manner is super critical. We have over 19,000 engineers that contribute to software. A lot of it internal, but a lot of it external as well. So we have hundreds of community managed projects where we contribute. We have hundreds of our own projects where we contribute. We also play this open innovations. Also plays a very key role in the company success. These community led projects, not just benefit Intel, but it benefits everybody. And the more we more inclusive, the more diverse feedback we get on these projects, the more it benefits the community and us as well. We also look at like, you know, how do we champion the code of conduct? You know, such as contributor covenant set up by Github and CNCF and all. So those are all critical element. And I would say the last point that is super relevant for us is how we are contributing across every layer of the global technologies to act. And I would say in that sense, Intel is very uniquely positioned being a silicon company. We can tune down to the wafer level and all the way to the application that is running in your application. So within my team and across Intel, we talk about how we are doing silicon to server less messaging that is relevant for the developers. So we contribute to entire layer of the stack and making it much more impactful and relevant for the customers. Thanks for sharing, you know, this and you're right. Also, we kind of live in a software driven word where everything is software defined. Now let's talk about this project open federated learning or open FL a framework. Talk a bit about what is this project all about. And also if you can share what is the origin of this project. Yeah. So if you think about federated learning, let's understand the concept of federated learning, why that is required. Now, if you think about any global problem that we need to solve, let's say you want to build a model that is really required to do the cancer detection in brain, for example. And that's a real example that I'm talking about. So this is where Intel was working with Penn medicine and identifying when we initially built the model, that model was built with the three states, California, Massachusetts and New York. Now, this is a global problem. If you were to deploy that model in a rural part of India or somewhere in Africa or somewhere else in the world, how would that model work? That is the problem exactly that federated learning is solving. Now the challenge over there is because of the regulatory environment, HIPAA and you know, all these GDPR compliance, these hospitals cannot really share their data. They said, take my data and go build the model and then give me the data back. It doesn't work that way. So essentially what federated learning enables you to do is you have an aggregator and then you have these multiple collaborators. You send the model to the collaborator. The collaborator runs the model in their own compute. They generate the model and send it back. So the collaborator is sending the model to all the aggregator. Sorry. The aggregator is sending the model to all the collaborators. They run the model. They send it back and then the accuracy improves and then that model is then sent back to everybody. So the security and the privacy of data is super important. Very much stays with the collaborator, but the model is the one that is exchanged back and forth. And that's where, you know, your concepts like confidential computing comes in. But how do we make sure that is data is fully protected? So that's the whole premise of how Intel was really working with University of Pennsylvania, VMware, and flower labs on how do we build this? And that is the project that was created out of Intel labs and then eventually contributed to LFAI and data foundation. See, you folks are, you know, contributing, you know, open federally learning to LFAI and data foundation. Talk a bit about what it will look like. There is it going to be a project within LFAI data foundation? What will be the structure also? There are no open source community. There are open source communities. So we'll also talk about what kind of community you're looking around open, you know, FL. Well, first of all, it's a project as part of LFAI and data and there is a GitHub repo. So if you go to github.com slash secure federated AI, there is a open FL GitHub repo over there. And it truly follows all the governance model that is proposed by LFAI and data. There is a contributing.md. There is a governance.md and all of those standard things that you would expect from an open source project. So today you may be just a consumer of the open FL project, but tomorrow if this becomes critical for your organization in a very classical meritocracy based way, you may want to rise up to be a maintainer of the project. And boom, you know, you can just look at that governance.md rise up, become a maintainer of the project. And honestly, this is how when I was talking to the open FL team, I understood VMware got involved because, you know, if you think about generative AI, those models, large language models are trained using 175 billion parameters. If you have to ship that back and forth between the collaborator and the aggregator that we were talking about, it requires a lot of bandwidth and the security element of it. And that's sort of the functionality where how do you ship those large models back and forth, was scratch their own itch element for VMware. And that's the whole element. So it's both a project and the GitHub code base where people can look at it. And they're very simple. Like, you know, if you can go to the demo directory or examples directory, you can really quickly fire up, you know, how, what the feel is. So there's very easy way for you to get started with open FL as well. You talked about VMware getting involved. Can you also talk about what kind of community you're trying to build around this project? Yeah, no, absolutely. So I mean, if you look at, you know, Penn Medicine is super deeply involved. I mean, the initial study that I talked about, that was done around three states, but then we expanded to 71 different sites. So we were really working with, you know, a lot of these medical institutes all across the world. So from three states just based out of the United States versus the 71 sites all across the world, that truly really allowed us to make this data that much more diverse. So now you can start thinking about what kind of communities we want to build around it. We want to work with the academics. We want to work with the medical researchers. You know, we are bringing the concept of federated learning and the whole concept of federated learning, if you think about it, was created by Google when you are typing your text message, you don't want your text message to be sent back to the server and that's where it was leveraging the whole, they published a paper of federated learning. So really academics, medical institutes, you know, companies, you know, who really see the value in this, who really value that the data cannot leave the jurisdiction and also anybody and everybody, developers, you know, how can they improve their entire developer experience, getting started experience, all of those are really valid folks that we would love to invite and participate and I would encourage, you know, literally it took me half an hour to just say pip install openfl and it installed the thing for me, then I have the FX, the CLI, which got me started. So I think I would just say, get started with it, open source is all about scratching your itch. So if you can say, you know what, this is what I don't like about it, submit a PR. You know, very well said open source is scratching your itch and once when I was talking to Linus or Greg Covertman and they're like, yeah, we do like a contribution but we don't like contribution for the sake of chatting. We want selfish users who are also using the same code base because that will make it sustainable. Otherwise, once you drop the code base, you're done. I don't even care about that. So which brings me to another point is that when we do look at today's open source word, there are two kinds of open source. First, you know, of course, a lot of companies, they put their code base in neutral foundation, like Linux foundation, which breeds a lot of confidence among your potential competitors. They don't worry about tomorrow, you'll change the license and suddenly log them out of it. At the same time, there are a lot of companies who still do a lot of open source but they keep control over the code base. Contributing has its own benefits because you have a lot of your community but you also lose control. You cannot direct the flow because that depends on the community or the whole ecosystem around their projects. So talk about what does company like Intel gain by not essentially releasing, but not only releasing but also putting their code base in foundations like Linux foundation. Yeah, I mean, I think what I would say there is the concept of Nora, right? No one right answer. And also people do open source, how it matters to them. A couple of weeks ago, I gave a talk at scale, which is the biggest community run conference in North America, about 3,000 open source developers and I was talking about fostering an open source culture. And one of the key elements over there is that the open source that you're doing, it really needs to be tied back to a sustainable business goal. That is super important. And as I mentioned in the beginning, one of the main reasons we contribute code out there is because we want to see that diverse community helping us make it better. It is the moral imperative for us that we want to create that horizontal competition where we create an equal leveled playing field. We want to break down those world gardens where one or two vendors are really defining and controlling what that agenda looks like. We want to enable that more diversity, more inclusion across the industry and more partners to make that solution that much better. Now imagine if that open FL case that I was talking about, if we were only running that within those three states, that model would never be successful. That's why bringing that richness of data across all the world is so much important for us. And as I say, a rising tide lifts all the boats. Sure, that's how we see it too. The reason LF becomes a really good home or any foundation for that sake, because it's vendor-neutral. As you mentioned, there is no copyright trademark issue. It's a neutral governance model. It allows folks to join in. That's exactly why University of Pennsylvania, VMware, Flower Labs, they are all joining the part of the technical steering committee because they know they will have an equal say in the vice. Sure, the project was originally created by Intel, but by being as part of the LF AI and data, there is more than 20 years of legacy of LF doing things the right way. So I think that is a part that really excites me. That is where Intel is part of what? Almost 800 foundations and standard bodies. There is a reason why we are so deeply involved in open source for over 20 years. And we believe that is the right model. We believe that is the right way to engage with the community. And we have contributors. We do job would carry water work. We are part of the sake leads. We are part of the foundation boards. So I truly believe that is the way, how we make the world a better place. Arun, thank you so much for taking time out today. Not only talk about this project, but also, I love the way you talk about the low-law organization Linux foundation of low-law computer like Intel by putting their own code base into open source. I love all this discussion. Of course, we talk about this new project as well. And I'd love to see you again soon. Thank you. Thank you. I really enjoyed it. Thank you for the opportunity.