← Lead With AI
Lead With AI
Jarvis in Your Laptop: The Decompute Revolution Bringing AI Home
Jarvis in Your Laptop: The Decompute Revolution Bringing AI Home
Loading summary
Transcript49 lines
- [00:00]
Dr. Tamara Nall
What if your AI didn't live in the cloud, but on your device? What if you owned it, controlled it? No subscriptions, no surveillance, just pure private autonomous power at your fingertips. Today on Lead with AI, I'm joined by Hina Dixit, Founder and CEO of D Compute, a company that's rewriting the rules of AI infrastructure for both businesses and consumers. With D compute for B2B and Blackbird for B2C, Hena's bringing AI out of the cloud and into your hands where it's faster, more secure and finally under your control. This isn't just decentralized, it's intelligent infrastructure, adaptive, self running and built for the edge. If you're curious about what's powering the next generation of apps, research and secure AI experiences, you don't want to miss this. Let's dive into the mind of a builder who's reshaping what's possible. Let's get into it. Welcome to lead with AI. I'm Dr. Tamara Nall. In each episode we will take you behind the scenes with visionary leaders shaping the future of AI across public and and private sectors. Join us as we explore groundbreaking projects and innovations that are transforming industries and making a real impact on people's lives. Let's dive in.
- [01:37]
Hi everyone, My name is Dr. Tamara Nall. I am the host for Lead with AI and today we have someone so exciting, we have Hina Dixit from Decompute. Hina, welcome.
- [01:49]
Hina Dixit
Hi Tamara. It's a pleasure being here. Thank you so much for this kind invitation. Really looking forward to chatting with everyone today.
- [01:56]
Dr. Tamara Nall
Absolutely. We are so excited about learning about you and Deep Compute and all of the great capabilities that it can provide for our listeners. And I always like to start with what I call the spark. So tell us a little bit about who you are at the core core and what problem were you trying to solve when you created D Compute?
- [02:22]
Hina Dixit
Sure, that's a very well asked question. Honestly, not everybody puts it this way. So I am the CEO and co founder at D Compute. Before this I was in venture industry. I was a partner at Microsoft's venture fund. I like their AI investments AI thesis. Also invested in infrastructure companies. Before that I used to head AI investments at Samsung's venture ARM called Samsung Next and I invested around 15 plus company before like you know Jenny I was Genai and I have like around six plus like you know, great exits. By God's grace, I don't think like I did anything and my job was to just support the founders and founders did it. It's just that like I still feel so proud of all of them. And before that my background is pretty technical, so I was at Apple for around eight years and before that two years at Symantec. At Apple, originally in my masters I started with computer networks and then I moved towards security and from cloud security I moved more towards AI as I grew and I started taking on these different AI oriented projects, be it something related to CVML or something related to performance or like bringing software 2.0 paradigms into Xcode like in some way, right? So started taking a lot of projects there and then like I grew up as, as they say, right. And became more of a leader there, right, like tech lead and started leading all the projects like you know, across different organizations, across different like teams, different products. And I had my breakthrough moments where I felt like, okay, like I know, like in AI and I know Apple very well now. Like I know the DNA of the company, I have the DNA of Apple in some way ingrained but at the same time like I was never exposed to the worldly problems, right? Like when you are at Apple you're only thinking about what's best for their consumers. You're only thinking about like the infrastructure and ecosystem that they have. But then there are problems that exist beyond that. And that's why like, you know, I stepped out of my comfort zone of being like, you know, a technical expert into moving into like, you know, venture venture capital capital, like, you know, industry, which was like kind of a leap of faith, honestly for me because I had a very well settled career at Apple. I love my job, I was getting recognition from everybody, right? Like, so I think at that moment like I felt I have to build something where people can use it. And if I want to like, you know, build something which is general enough, I need to step out of the Apple's ecosystem and learn more about different business problems. And that's how my mindset was always that hey, one day I'll start my own company. And then decompute happened because I had this moment of clarity. And the moment of clarity came when I saw so many founders, they were trying to raise funds from these big incumbents and cloud provider companies and their reasoning was, hey, we want to get ahead but we do not have enough, you know, resources to compute. And they were like in line to get access to Azure or GCP and like, you know, Nvidia or Oracle, right? Like they were not getting enough compute. And I felt that that was slowing up the revolution, right? Like the adaptation became much slower and that's where when it kind of strike me that, okay, like, you know, we have so much of untapped potential in these, you know, laptops, your Mac Minis, macros, Windows machines, Windows servers, AI PCs, which can be tapped into computing as well for AI. So that's how decompute started. So I started working on that thesis and then I continued building on top of it.
- [06:14]
Dr. Tamara Nall
So that's the story that's amazing. It's about access for everyone. And that's what I love most about this podcast, is that it really highlights those solutions that open up opportunity for people, giving constraints and obstacles. So that's, that's, that's amazing. So now talk to us about, like, the Holy smoke moment. So what's the jaw drop moment? When a person first experiences your product and realizes that this changes everything.
- [06:44]
Hina Dixit
Okay, so I think that moment would be Tim Tally, like, you know, experiencing Blackbird for the first time. Tim Tully is basically like, you know, one of the lead investors and pioneers in AI, right? Like, he actually, like, you know, is one of the investors at Menlo Ventures. And he was blown away. Not just him. When we met Guido from Edwards and Horowitz, he was blown away as well. And both their reaction was, where, when are you loading your mittimars? How can this be so fast? What are you doing under the hood? And, and sadly, you could say, right, like, because we need to maintain the ip, we couldn't like, share the ins and outs of things, like how it was working. And in the end, like, you know, Tim was like, hey, congratulations. I think this is, this is amazing. Like, you know, I don't know how you're doing it, but this looks amazing. And I think that was the holy Smokes movement, like, for all the founders and that. Like, you know, there are these people who are recognizing it and given the.
- [07:45]
Dr. Tamara Nall
Fact that they see so many investment opportunities and then they're in awe. And having that holy smoke moment is amazing. Now, I know that you have IP and you can't share it, but what can you share? Like, if we were to, to open up the hood or open up the brain of decompute, what you know, in layman's terms, what will we find?
- [08:06]
Hina Dixit
So if, if we were to think about this, think of this, that it is more of a local omni science drive, right? Where you can instantly deploy any model on your own device. You can train it on your own personal data and generate anything that you imagine and you never have to touch the cloud. Now you take total control of your data, of your AI of your skill set, which is the most important thing here, right? Like, because you can actually teach your model how to like, you know, generate right answers or do certain tasks for you. But also zero latency. Because with lasertune, which is the powerhouse behind Blackbird and lasertune is one of our proprietary technology, we have made fine tuning as fast as rag. Basically you give it a hundred page of file, it could fine tune within 10 seconds, which is amazing. There's so much of creativity that is there to unleash here. And we are soon launching image generation as well. On device, you can basically generate Ghibli or any type of art styles like locally on your device. And you don't have to pay like, you know, these big incumbents to do that. So that's the generic like, you know, thing that we are trying to do, making AI like global.
- [09:30]
Dr. Tamara Nall
And so does that mean. Because what a lot of LLMs is pulling from all these public sources and there of course with some people is a concern about security. But it sounds like with your offering you don't have to worry about that because it's all here locally.
- [09:48]
Hina Dixit
Yes, that's.
- [09:50]
Dr. Tamara Nall
Now talk to us about one of your customer experiences that really blew their mind. You know, we've had guests where they would talk about just some big like, wow moments where a customer was using your product and they could see it for themselves. Talk. Walk us through that story.
- [10:08]
Hina Dixit
Sure. So one of the customers that we had is. So they are basically like in finance industry and their job is to like, you know, constantly monitor all these like, you know, different reports, right? Annual reports, SEC filings that come their way and constantly like, you know, monitor the movement of the market and also predict it. So it's not only monitoring, but it's also predictive intelligence that is needed to be happened like in real time and how fast like Laser Tune and Aurora can make it happen for those like, you know, consumers. That's mind blowing, right? Like so because we can actually parse the entire like, you know, SEC file, which is a very complicated multimodal data, if you think. Right. Like it has tables, it has images, it has graphs. There's so much of like textual like nuances in it, right. So there's so much math in it, right. Like, you know, and the math could be very confusing, right? Like because the same numbers are being used in different context and then they could also mean different things. So now like you're combining like a powerhouse here, right? Like of a lot of multimodal data, but also reasoning on Top of it, right. And that's where we come in. Because if you were to do this just with inference, right, like just host a very small model locally and give it like you know, some sort of like you know, file and then ask it to do SWOT analysis or like you know, ask it specific question that. Okay, like who are my top 5% stakeholders or how much stake does like you know, Ajay Khipuri like holds in Nvidia. Right. Like usually like biggies like perplexity or Jubilacy. I cannot pull that out. And those people are on cloud but we can do that on device for them. So suppose if you got your hands into like a very personal like you know, private financial document and you're trying to quickly parse it, get these numbers quickly and get reasoning on top of it, you can use like Blackboard for it. And that's the aha moment that I think a lot of like our customers are witnessing at this moment.
- [12:11]
Dr. Tamara Nall
Got it? No, that's amazing. And so is your product for both retailers slash consumers and B2B. So who is your ideal customer?
- [12:24]
Hina Dixit
So we have like a product called Blackbird and then there is a product suite called D Compute. Right. So D Compute basically like has a platform which is more, you know, more built towards compliance and like you know, optimization for enterprise adaptation. Right. Whereas like Blackbird is our more consumer and pro consumer facing product. So that's how like I think we see it. We want to sell it to enterprises who are trying to adapt AI quickly like at a lightning speed. And then like you know, we are also targeting people who are very privacy oriented and they want to own their AI locally on their device and they want to build their own agents. So these two customers are where like segments where we thrive the most.
- [13:13]
Dr. Tamara Nall
Perfect. So D computer for your more B2B enterprise customers and then Blackbird are for your retail consumers.
- [13:22]
Hina Dixit
Yes.
- [13:23]
Dr. Tamara Nall
And then do I get, do I go to Blackbird from D Compute or does Blackbird have its own access?
- [13:30]
Hina Dixit
So you basically go to like our website Decompute Run and you can download Mac OS right away for Windows. We are launching next Monday. So upcoming Monday we'll have like you know, Govdash for Windows and Mac OS both for running and training models locally.
- [13:45]
Dr. Tamara Nall
Perfect. So by the time this is released everyone Blackword will be available. So you can go to D compute.ron for that. Now let's talk about the ethical and you and I were kind of talking about this a little bit before we started the show. Let's talk about the ethical Crossroads. I mean, what you're talking about is that people have access to very powerful tools. And so what ethical lines are you watching? And how do you protect the human in this loop?
- [14:13]
Hina Dixit
Yeah, sure. So first thing first, right? So ChatGPT recently announced that the context window has become forever for them, right? Like, it's, it's humongous. Now, a model that is deployed can remember your entire conversation in your past life, right? I think we as humans deserve a break. Don't you think so? I think not every moment can be a moment of perfection for us. But everybody, like, everybody who has made a mistake in their life in a human world can go and can make mistake, but can forget, be forgiven and move on. There has to be a similar journey for people in the AI world too. And we call it right to forget. And that is something that I think, like, you know, the world is not thinking about here, right? Like some parts of the world, like Europe, right, Like, or the euro region, like, in general, they are thinking about this. But I think, like, you know, whatever data that you put on, whatever picture that you put on can actually be manipulated and you could actually, like, hackers can get your data out if it's on cloud. It's very like, you know, prone to hacking in some way. But then how do you, like, even trust, you know, people with these private moments, like, you know, very private, like your images or very private financial kind of data, right? Suppose, like I'm, I'm just a normal person. I need, like, to look into my finances and I want to, like, feed in my bank balance statements and I want to ask questions. He, like, how do you think, like, I should budget this? How do you think, like, what, you know, expenses do stand out the most. How can I cut them, right? Like, ask these simple questions. And I don't think, like, that's possible if, you know, we do not do the work that we are doing here at Ecompute. And that's where we come in. We know that, okay? Like, not every moment needs to go on the cloud. Not every moment or every part of your data and life has to be on cloud. And that's where we are forming a wall between the cloud and between the users so that our users can feel secure and they can still use AI in the closeness of their own device, in their own house. You know, zero contact with cloud.
- [16:29]
Dr. Tamara Nall
Yeah, no, that's amazing. So I just had my aha moment while we're talking, because I was thinking about the times where I might use ChatGPT or, you know, Cloud or whatever I'm using. And there are some things that I do not upload because I know, you know, anybody can pull from it. I don't know necessarily how it all works in terms of their pulling it, but they can. And so I still find myself doing a lot of well this is automated as I can in terms of programming my model. But you know, I do find myself doing the work myself. The analysis of like particularly with taxes and you know, with my having different businesses, my nonprofit and trying to segment those. And it sounds like, you know, Blackbird in particular for me as a consumer could very much help with those. That analysis or that research that that is private to my own life that I don't feel as secure about.
- [17:22]
Hina Dixit
Yes, absolutely. And that's where we come in and that's why we say okay, like let's help you build these agents on the go. So right now we have six type of agents. There's a financial agent, there's another agent which does your legal documentation. Then there is an agent for doing analyzing, you know, really like long format documents or research work. Right. These could be hundreds of pages like books you are reading. And then like you know, there is one for personal conversation like a meeting which you can revisit and ask questions about that. Hey, like what were my task items like with this meeting? Who were the, you know, who were the top stakeholders in this meeting? Right. And then we have like a very generic one and one for the general coding one as well. Like you know, just for seeing that how well like even small models can do like if hooked onto a single repository or like you know, a couple of repositories.
- [18:14]
Dr. Tamara Nall
Now can I connect to the cloud if I want to or there is no connection at all.
- [18:18]
Hina Dixit
So right now we don't have connection but if somebody wants it. So we are planning to have a cloud connection and then having an automatic filtering which can understand your complexity of your task and can also understand how we can be the cookies in the cookie less world for AI while maintaining your privacy. So we keep your data private, but we get all the research, all the input from the web, from the cloud and we give you the best answer locally. That's where we are headed from the Blackbirds side or the POV that we want to be that filter between the people and the cloud in general. That's the plan.
- [19:03]
Dr. Tamara Nall
Great. We have a couple of developments that are coming out soon, but let's think about the big future. So let's Fast forward to 2030. We're going to what does the future powered by Blackbird or D Compute, both have in store for us, for the world.
- [19:20]
Hina Dixit
Well, we imagine that every device basically becomes an AI powerhouse in the future. And AI companions can run locally on your phone, on your car, your laptop. Jarvis like Ironman, right? Like, except that Jarvis becomes local or it controls all your devices in your house, but then it's all local like resides locally in your house, just in a small like you know, laptop or a small Mac mini. And then your data never leaves your hands because your compute never leaves your device with the compute. So that's how we are envisioning this. And then like eventually we want to not only like empower people but also empower the enterprises where their dependence becomes lesser and lesser. Right? Like they can make more profits just because they can push some of the computer locally on device and they do not have to pay to these giant incumbents for running compute. Or they compromise on their users privacy by sharing the data with these cloud providers. So we become the new future of compute. We are not saying that cloud will go away, but at the same time when it comes to serving to the consumers, we will be the people you'll come to for, you know, helping out. Because we are focused on people's first priority, which is their privacy and security.
- [20:41]
Dr. Tamara Nall
And you know, that kind of shakes up the revenue model for some of these cloud providers because they very much take what they know about consumers and you know, analyze it and then potentially sell it to other marketeers. So that kind of shakes up that whole revenue stream and they have to start thinking about that.
- [20:59]
Hina Dixit
That is absolutely true. And that's where I think we want to take this too because eventually, right? Like not everything, not every query needs to go to the cloud, right? Because a lot of these queries we are actually showing we can process like entire CC filing which is like 200 or 100 pages at least. Such a complicated data like for a very small model residing on a laptop, right? So eventually we see a world where your COMPUTE usage on the cloud will become like, you know, less than 25%, right. Like for most of the tasks that you want to do, they can happen locally on your devices. And if that's the case then just imagine like how much power does this give to enterprises who are selling or these developers, right? Like who building for their consumers, right. And they can actually, they don't have to pay the cloud providers anymore, they can actually directly ship it to their customers and their customers can control, right? Like whether they want to use the cloud or they want to Use like, you know, their own local device for this?
- [22:00]
Dr. Tamara Nall
No. That's amazing. Now, one thing that I do love to talk about is, you know, our listeners are here. They want to know how can I put this in action today or tomorrow? So talk to us about one thing that our listeners can try build, explore this week with Blackbird or decompute to really turn it into action this week.
- [22:24]
Hina Dixit
Sure. You can download Blackbird right away and you can start using it to code any project. Or you can also, like, you know, use it for any legal documentation. You can build your agents on any of these six sectors, which is legal. There's another one technical, this third one called finances. And there's one which does, like, more of a generic work for you. And then there's research. There's another one called meetings. So it lets you, like, record all your meetings locally on your device, no meeting, going to the zoom, and then, like, it will always remember your meetings for you.
- [23:00]
Dr. Tamara Nall
Wow.
- [23:01]
Hina Dixit
And the notes as well, right? Like the meeting notes and stuff. So you could always, like, put your notes and say, hey, like, can you summarize my notes for me?
- [23:09]
Dr. Tamara Nall
No, that's great. Well, I have thoroughly enjoyed learning about both Blackbird and D Compute. And if there was one phrase I could come up with, it's like power to the people. Like, it's bringing the power back to have it, you know, secure on your devices, in your home, in your ecosystem and not be as dependent on the cloud. So what a phenomenal technology, Hina. And to see you at the forefront of this is amazing, particularly given your stellar background.
- [23:39]
Hina Dixit
Thank you so much. Thank you so much.
- [23:41]
Dr. Tamara Nall
So now we're going to go into our rapid fire. So I have four questions to ask you, and then we have something that we call from one genius to another. So my last guest has a question for you. So I'm going to ask you that. So first, what's the most overrated tech trend?
- [23:58]
Hina Dixit
I think the most overrated tech trend right now is inference exhilaration, I think. How fast do you think it should get? Where would it go? Right? Like, why do you need it to be so fast for consumers, for your research work, for trading for, like, you know, predicting any risk in the markets, like, those kind of things where, like, market movement is very volatile and very rapid. Yes, I do understand, like, you know, it's very useful. But after a point of time, like, a user has like, only a certain, like a definite reading speed right where it feels comfortable. So I feel like that has to be kind of like understood from a User's perspective too. Right. Like what we're eventually shipping to them.
- [24:46]
Dr. Tamara Nall
Yeah, right. The speed expectation. Okay, what about the most under hyped AI breakthrough?
- [24:52]
Hina Dixit
I think the under hyped AI breakthrough, I think that would be us at this point because see, like, people know on device is coming, but also like, you know, people are more hyped up about like inference acceleration or like agent creation, but they're not thinking as much about the privacy and security. I think Euro region has done much work in that area. Then I think us at this point and I think that's where we need to start thinking as being the leaders in AI. Right. Like us has always been leader. So how can we still like, you know, give people right to forget, but at the same time, like, you know, continue building AI together?
- [25:31]
Dr. Tamara Nall
Got it. Okay. What about a book folks should read?
- [25:35]
Hina Dixit
So I really like that will never work from Mark Randolph. And I like it because it not only like, teaches a very ideal life for a founder, and like, you know, it, it becomes your, like, you know, a companion for your soul as a founder. There are a lot of like, you know, personal life highlights and snippets in that book and I think, like, I really, really enjoyed reading it. So I would recommend it highly to any founder out there.
- [26:01]
Dr. Tamara Nall
All right, perfect. And then the boldest AI prediction that.
- [26:06]
Hina Dixit
You have, the boldest AI prediction is us. Cloud is not the future of AI. You know, it's going to be living on device.
- [26:15]
Dr. Tamara Nall
I love that. And then from one genius to another, your question is, if you could write a rule that everyone must follow for one day, what would it be?
- [26:26]
Hina Dixit
So for me, my question would be that, you know, if everybody can be sincere in their journey every day, right? Like, what would we see, like, and witness a very different world? Because right now I think we have come to a very hustle culture. I'm seeing like, people pick up these ideas very quickly, like hustle quickly and then move on. But then, you know, I, I kind of like don't see people putting sincere efforts in their life every day these days. And I think I kind of miss that part. I think I want to see that passion again in people. And I've seen it in some founders. Don't get me wrong. Like, it's not about the general world, but it's, it's about like, you know, it's more about changing the direction because we have to be the best in AI and how can we do that? We can only do that if we are sincere and we start thinking of this from more holistic level rather than thinking of this being a money making vehicle at this point.
- [27:25]
Dr. Tamara Nall
I love that advice for the soul. So hina, thank you so much. I've enjoyed this discussion and I'm sure our listeners are now just as excited as you are about Blackbird and Decompute and how it can really kind of like change how we think and analyze and use AI in our everyday lives. So again, thank you so much for being here and I thoroughly enjoy the discussion.
- [27:47]
Hina Dixit
Thank you so much. Tamara, you're such a kind person. I was looking at your background, phenomenal background. Very inspiring to be here with you, like such a great moment for me. Thank you so much.
- [27:57]
Dr. Tamara Nall
Thank you. Thanks so much.
- [28:00]
Thanks for tuning in to lead with AI. I'll see you next time as we continue exploring the cutting edge innovations shaping AI across the public and private sectors. Until then, keep leading with AI.