back to All Episodes

Combining AI and Robotics

Mahesh dives into building a venture-backed AI Robotics company, sharing lessons learned along the way.

Listen to this episode on:

Spotify
Apple Podcasts
YouTube
RSS

✍️ Show Notes

Mahesh Krishnamurthi, co-founder of Vayu Robotics, gives an inside look at the journey of creating an AI-powered delivery robot. Important questions we discuss: how to know when you have product market fit, how to grow a talented team, and how to know when you should raise money.


🔗 Check out Vayu Robotics: vayurobotics.com


🔑 Bytes:

  • Prototyping should always start with off-the-shelf components to minimize cost and speed up development.
  • There's a fine balance between confidence and arrogance when it comes to pitching investors - lead with humility and quiet confidence.
  • Be cautious if you're raising capital before you hit the flywheel of product market fit.


💬 Full Transcript

Vignesh Rajagopal (00:00) Welcome back to another episode of Fika Bytes. This week, we're talking to Mahesh Krishnamurthi, one of the co-founders of Vayu Robotics. Now they're combining AI and robotics design to create these autonomous solutions that are going to be low cost, scalable, and environmentally sustainable. And you know, AI and robotics, they're not just buzzwords that are being thrown out here. Vayu actually has the first product that's already out on the roads. It's called the Vayu One. It's a last mile delivery robot, fully autonomous, has its own loading, unloading mechanisms. It is just a sight to behold. And, know, it's not just this product that Vayu is building there. As Mahesh puts it, they're building the nervous system for all robots. So taking this technology, this research that they've done and applying it to other types of robots, for example, they just released a four legged robot. And, know, it's not easy to build this technology. So Mahesh dives into building a team, raising the capital, the personal roadblocks that he met along the way in order to make Vayu Robotics into the company that it is today. We recorded this in July of last year and since then, Vayu been making all sorts of headlines, Forbes, NASDAQ, TechCrunch. They got a huge client order. It's really cool to see the progression that they've made since we recorded this episode. So let me know your feedback as you listen to this I would love to know what you think so I can continue bringing these hardware founder stories out into the world. All right, enjoy the episode. Vigs (01:20) All right, we're live. I've got Mahesh here with me from Vayu Robotics. Mahesh, thank you so much for taking the time. I'm super excited to talk about what your company is doing in the AI robotics space. You guys are offering software package, hardware package, and then like a combined solution off the shelf as well. So tell me a little bit about the company and how did you start Mahesh (01:38) Yes, So, Vayu Robotics was started maybe two and a half years back and, you know, I've started it with a few of my friends. You know, I like to say To start a company, you need to find people who you can go the journey with, not just during the ups, but also during the lows. So I was very, very fortunate to find Anand Gopalan, who's our CEO. And he was a former CEO of a very famous public company called Velodyne Lidar. So he's one of my partners in this. And I also have Nitish Srivastava. He was my former colleague at Apple. And, you know, one thing about Nitish is the combination of very, very humble and very, very smart people is very small in this world and he's one of those guys. So I'm really fortunate to have got to work with both of them and in building this company. So what we do at Vayu is basically we are building this delivery robot. You know, when we started out, we've all worked in autonomy for a while. Maybe collectively we have around two decades of experience trying to solve this very tough problem So we had a very good understanding of what works and what doesn't work. So we came together and then figured out with whatever we know works, is there a product application that could be a good business model? So then we came up with some concepts for our products and what we ended up with was a delivery robot, which is not very slow, but at the same time, it's also not very fast. So it has more time to plan. And also it's not very heavy because if you want to increase safety, the biggest lever that you have is the weight of the robot. So we believe that given the state of autonomy today, if it had to ever into the public market, you start out with something which is slower and also goes at reasonably low momentum. So what we decided was we wanted to mimic the momentum and energy of a bicyclist. So the robot that we wanted to build was something that can go at Usain Bolt speed, like basically a bicyclist speed at 15 miles per hour. And it can carry around 200 pounds, including the payload. So 100 pounds of payload and 100 pounds of robot weight. So that's what we are building. We are a team of around 15 people. I really believe in the power of small, focused teams. And each one of us here have a very strong experience in one area or the other. So there is a knowledge gradient between any two people. And we are very focused on bringing this to life. Vigs (04:25) Awesome. And what's your kind of ideal customer profile? Who are you looking to make this product for? Mahesh (04:31) Yes, so we have a few customer profiles. One is for grocery delivery, which can be more scheduled and go from like a dark store to an end customer. So that could be one customer. So we would lease our robots to them. And it would be a model where you pay per month for a robot. And this would take care of a lot of deliveries when there is peak demand. We also have other customers, like people who want to deliver food. There we go directly to restaurants and have a deal with them that during peak hours you can use this robot like a lease. But those are the two approaches that we have. And there are other peripheral applications, but this is the customer base. Vigs (05:18) Okay, nice. And you said it's been around for maybe two and a half years. Can you share a little bit about, you know, what is the current progress? Do you have prototypes or you have units in the field already? Mahesh (05:21) Yes. Yes, so we have built around 14 robots until each of them have been an evolution. as you know, probably building hardware is not a very straightforward process, especially with a small team. So we do have a lot of prototypes and every prototype is meant to answer a question that we have, which is unanswered. So we have built around five types of prototypes. And right now, the prototype that you see in the back is deployed on public roads and we are doing real world operations autonomously. So we have through multiple increments and we've come to this form factor in over the time of two and a half years. Vigs (06:12) That's incredible. I want to pick up a little bit that you said. You said each prototype helps you answer an unanswered question. Can you just elaborate what you mean by Mahesh (06:19) Yes. Yes, so typically when you do an experiment, you want to find an answer to some question which you cannot find an answer to with existing information. So that's how you structure an experiment. So for example, when you want to come up with an unloading mechanism, you would have an idea and you would think that it may work in the field. But when you develop it, maybe there are a lot of unknown unknowns when you when you design a project, there are known unknowns and then there are always unknown unknowns. So doing a prototype helps you gain some insight into the unknown unknown category. And once you get some knowledge there, you refine your thinking and then you iterate on the design. So that's what I meant by there are always unanswered questions that help you make progress along the way. Vigs (07:18) Yeah, I think the point of iteration is an important one, especially in hardware and software development. I'm going to oversimplify this, but in software development, you can just change a few lines of code and then you have a new iteration, right? And that is an oversimplification, I admit. In hardware, you often have to start from scratch. If you're sourcing parts, maybe you have to find replacement parts. So each of your prototypes that you built, what was kind of like your approach of reusability and improvements, just like the mentality for Mahesh (07:29) Yes, Yeah. Yeah. So, you know, as a small startup, you know, coming from a company which ships millions of products where we optimize for the quality and customer experience the most. The biggest change that I personally had to go through coming to a startup was we wanted to optimize for speed of execution. So that means that you have to sometimes let go of things like high temperature performance or reliability with shock and vibration and things like that. Just to, and maybe size or the form. So we have to make some choices which are optimized for speed, but not necessarily thing that would ship in millions of units. So as we go along this journey, we minimize the speed optimization and then add more ruggedness into the robot. So that's basically what we did along the way with every build. You know, like for example, this robot has a lot more design elements into it. The first robot that we built was like a cuboid, you know, just a very bare bones functional element. And we didn't have any custom technology. So when you want to build the first prototype, you want to take as much off the shelf technology and components as possible to minimize the risk. And once you do that, you can always add new technology to it, which gives you a competitive advantage. So what we did was we took two pieces of technology. Vigs (09:06) Mm -hmm. Mm -hmm. Mahesh (09:17) which are proprietary to us. And this particular robot has both of those integrated into it. Whereas our first one was mostly off the shelf optimized for speed. Vigs (09:28) So proof of concept was the first one and you kind of took everything that was already available on the market. And then I guess, can you tell me about the transition from that to developing the proprietary technology? What made you decide to do that? Or was it always the Mahesh (09:30) this. Yes. Yes, so our main hypothesis was that when you combine deep learning with currently existing cameras and vision -based technology, we would be able to get a neural planner that could plan a path at a much lower compute overhead. So that was our hypothesis to start with. And we tested it with our first prototype to just test that idea. And once that worked, then we started adding more functionality on top of it, like ability to deal with corner cases and then ability to do automatic unloading, ability to carry more payload, ability to go longer distances. and things like that, and ability to do this planning much faster in corner cases. So those were the things that you start very simple and then you add more complexity as you go along. Vigs (10:38) So that first experiment was that kind of like, you know, you and your co -founders, you guys were like, if this doesn't work, you know, we have to abandon it, but hey, if it works, we have to move forward and build a business around it. Was that the understanding? Nice. Mahesh (10:49) Yes, yes, yes. So there was a technology risk to start with. So the first round of prototypes were focused on minimizing the technology risk. Now, we are looking at market adoption risk. If a device like this exists in the world today, what would be the market adoption? Because as you know, the delivery robot proliferation has not been super high. So that's the thing that we are trying to explore at this stage. Vigs (11:21) Super cool. And I want to talk about you and your co -founders and what kind of was your background before you decided to start this company. You talked about having some experience with larger companies and the similar technologies. So what was it that brought you all together to work on this? Mahesh (11:24) Yes. Yes. So, you know, like I mentioned, we are three of us and Anand, I met him. So I used to maybe I'll give you some background about my journey and then I'll tell you where I intersected with my co -founder. I started off after grad school working at Intel labs on a silicon photonic transceiver, which was, which is currently used in a lot of data centers and helps you upload live video fast. So it's a technology which is used. Vigs (11:48) Yeah, let's do Mahesh (12:07) and high speed data communication. I started working on the group and it was really small. So it gave me like a zero to one experience. And then after that point, I moved to Apple for around seven -ish years, where I started off working on the display team, shipping very high volume products like iPads and other large display form factors. There I learned the importance of rigor and paying attention to detail and also the effect of teamwork, very tight teamwork. I learned that there. And when I was at Apple, I moved to the self -driving or SPG as they call it for the last five -ish years. And that's where I met my first co -founder, Anand Gopalan, who was supplying lidars to Apple. So I knew him, I was his customer. And that's when I met him and he was, know, that from then on I used to keep in touch with him. And my other co -founder is Nitish. He is a deep learning expert. So he's from Professor Jeff Hinton's group and his background is in training neural networks, you know, coming up with new architectures for neural networks and his company out of grad school got acquired by Apple when I was in the SPG group. So that's how I got to know him. And then I used to like, he also used to work pretty late and then we used to meet for random chats at the coffee bar. And that's how I got to know him. When we thought of this idea, then I reached out to them and wanted to know if they would be interested and they were very happy to join forces to solve this problem. So that's how we met and that's our background. So my background is mostly in optical physics, sensing, hardware system integration and Anand. came from a business background where he was the CEO of a public company and Nitish has a deep learning background. So I think we have a good combination of complementary skills with enough raw ingredient. Vigs (14:07) Yeah. Yeah. Good. Good power trio. What was? Mahesh (14:11) Yeah, it's yeah, go Vigs (14:14) I was gonna ask, what was that conversation like at the point where you guys decided to do this? Did you all still have full -time jobs or it was like you were looking for something to do? What was the mentality? Mahesh (14:21) Yeah, so yeah, so the I think I've always wanted to do a startup. Uh, like I said, uh, there is, I'll tell you the hardest thing for me personally to do a startup was, um, you know, after working in the tech industry for maybe 10 ish years, you get to a certain level of financial stability and, uh, to let it go is super hard, meaning, uh, it's very, uh, destabilizing. So I feel, I personally feel like that is the biggest, uh, roadblock I've had to starting my company and now having done this for three years I feel like I should have done it way back, a long time back because the biggest risk in my opinion is is not taking enough big swings at the ball. Like I use, I play sports. So I think if you, if you keep playing for the safe shots and, you know, try to be very conservative, I think that's a very big risk. And being in a big company somewhat pigeonholes you into a certain role where you're not able to expand into areas where you're potentially capable to doing and, you know, do more good for the world than what you could do in a big company. So I think that feeling was always there in me. So throughout my journey, I've been trying to, it's been like a prototype building phase, you know, learning the zero to one at Intel labs, and then trying to get a bit more riskier projects like trying our self -driving group and stuff like that. So. When I had to make the switch, it happened at a time where my group at Lyft got sold to Toyota and then I had to pick whether I wanted to stay back or whether I wanted to go join some other company. And that's when I reached out to my friends and wanted to know if they would be in a similar position and they were happy to join forces. wouldn't say, so I would say forming a team that you can trust and also ride the highs and lows with because there'll be a lot of dropped catches along the way right like startup is not a perfect situation so you need to be able to have that comfort where you're okay backing them even things are going bad so I think finding those people was like a you know step for me Vigs (16:53) Yeah, definitely. For the financial stability piece, you said that you should have done this a lot sooner. So how did you do it? You you had this financial stability from these big tech companies for so many years. And then when you decide to do the startup, I'm sure you had to convince, you know, obviously you have to convince yourself, have to convince wife, maybe parents about, hey, we're going to take this big risk. And you believed, you Mahesh believe in the big swings, but how do you get other people in your environment to be on board Mahesh (17:00) Yes. Yes. Yes. Yes. I think it's a very, very important question. I would say if you're married, your spouse is probably your first VC, meaning your first believer. They write your first check. And they have to believe in you more than you believe in yourself. So I feel like the support that you receive from your partner is unmatchable in this journey. They see you day in and day out. So they have to be backing you like an order or multiple orders of magnitude more than anyone else who backs you. So I think I was very, very fortunate to have that backing. So I would say that kind of relieves some of the burden but even otherwise even if you're not married and if you're you looking to do this I think There is a lot of companies which are very small these days and there is no, like its capital is actually not the limiting factor for startups in my opinion. It's more about finding product market fit and also solving hard technical problems. There are a lot of, like if you look at the dry powder available, there's a lot of people willing to deploy that capital. but they are probably waiting for the thing that would give them the return that they're looking for. So I personally feel like capital is not the thing that limits startups. So I wouldn't over index on it. That's why I'm saying that I should have done it a long time back. Vigs (18:55) Because you're confident that your solution is important enough, you would be able to raise the capital. Mahesh (19:00) I think because if you build something of value to people, I believe that the VCs will come instead of us trying to go behind what they are doing. I do feel that if we have conviction that if we build this and it will provide value, I do think the money will come. Vigs (19:22) I think that's a beautiful way to put it. like don't start out looking for the money, start out looking to add value, solve a problem. So great, great insight. Okay. So fast forward to, you know, a few years later now the company has been built. You said you've got 15 people on the team. Um, tell me a little bit about the, you know, like the, the soft skills side of it, as far as you had all this stuff you learned from a big company. Um, what are some of those things you brought in and then what are some of the things that you added on or you improved to have like a lean mean team Mahesh (19:29) Yes, yes, Yes. Yes. Yeah. Yes, yes. So I would say, I think the biggest thing is we try to help each other genuinely, not just with lip service. But the reason I'm saying that is, We, we, we realized that we win and lose as a team and there are no individual winners in a small startup. And the company is basically the team, right? Like basic without people, there is no products or anything that happens. So we place very high emphasis on the people aspect of things. We have a very balanced culture. And I think that is very important for people to do their best work. And. When I say we help each other, Usually there is a tendency to have very bright lines in what you own in big companies, which is important for you to have a very good organizational dynamic going. But here, when we observe someone having a hard time trying to solve a problem they are at, we actually celebrate that, failures are celebrated just so we try to optimize for the effort side of the equation more than the outcome. So when you do that, then people naturally feel like they want to give their best shot and take whatever comes out of it. Right? So I think that's something that we try to do on a day -to -day basis. We are not a 10 on 10 on it, but we strive to be a 10 on 10 on it. I think that is very, very important in a small stage startup. Yeah. Vigs (21:24) And then at what point at this, of this journey, you know, it was a three co -founders, you guys to start off. At what point did you decide we should start building a team? We can't do everything on our own Mahesh (21:34) Yeah, you know, like one of the things that I believe is you have to find people who are better than you. So that's I genuinely believe in that so that you can learn from them along the way. So even though we are a well rounded co -founding team, there are a lot of gaps in what we want to do versus the skill set that the three of us possess. We can somehow extrapolate in those areas, but we do value expertise. So that was the reason when we thought it would be good to have a team to fill those gaps. So that's why I said almost everyone in our team is a go -to person for something and there is a knowledge gradient between any two people. I know they are experts in what they do. So irrespective of what seniority level is there's always something to learn from one person or the other. So that's when we thought we identified, for us to realize the product vision, what are the skill gaps we have and who are the Mahesh 2 (22:36) people in the industry and where can we find them and can we find them in our network and then that's how we went and built the team. Vignesh 2 (22:46) Yeah, you just laid out a full blueprint there of how to successfully hire. And I really like the knowledge gradient part. Mahesh 2 (22:54) Yeah, I think it's very important to accept that there are always gaps in knowledge and there are always experts outside the brain because there is a bandwidth limitation on how much a human brain can hold, right? So going to experts will only increase the product quality. So that was one of the things that we always keep in mind. Vignesh 2 (23:18) I want to talk a little bit about your journey with raising capital and talking to investors, giving presentations, things like that. Any key learnings, key takeaways that you would advise to others? Mahesh 2 (23:30) I wouldn't say I'm any expert in this field. There is always a long way to go irrespective of where you are in this journey, but I don't want to share one thing which could be helpful for early founders. When you start a company, you always feel like you're not good enough for the role and you try to compensate for it by working really hard and doing lots of other things to close that gap. And the thing I think the investors would index on is confidence, the ability to communicate that we are subject matter experts, we know what we are doing, and we are super confident about where this could go. And one thing I have found is when you're early in the founding journey, you're always trying to prove to people that this is going to work. going to work and things like that. But I think there is a line, very thin line between being arrogant and being super confident. And this is advice from one of my people in my network who gave to me that when you pitch, when you pitch or when you share your story, you should always be in that gray zone between being super confident and arrogant. And the way it comes across is, here is my story. You can invest in it if you want, but if you're not, there's no issue. There are 10 other people waiting outside the door who would be able to put capital in you. And I think getting to that level is non-trivial. It's easy to say, but... to really internalize it and then behave that way takes a long process. And that takes only time, right? You have to believe in yourself. Like things which are easy to say in words are actually very hard to practice sometimes. So I think that's what I think founders could strive to do because they are really the subject matter experts. taking confidence from that and... making sure that you're not running after the next shiny thing, but genuinely trying to build something of value to the world and adding more or solving hard problems and believing that if you do that, there is a market for it. If you have that first principles type thinking figured out, then I think the capital would come somehow. Vignesh 2 (26:08) That's really good advice for any early founder who's thinking about raising money. What would you say for those people who they have not yet decided whether or not they should raise money? Because if I were to guess in your case, this is a very capital intensive company that you're taking on. So in your case, maybe that decision was more financial, but how would you advise someone to make that decision of go the bootstrapped way or go the venture route? Mahesh 2 (26:31) I would say, it's a very good question. If you can, I would say as a rule of thumb, I think it's always better to take as less as possible in the early stages until you hit a flywheel and then you optimize for growth. think that because a lot of money in the early stages results to lot of bad habits and your culture gets all, it's optimized for something else, right? I do believe that having constraints in capital actually is good. So that's why I'm saying that early on in the journey, you should try to take as low as possible. Even though in the news, you always see like this person raised X at YGV and that valuation. So you can get distracted, but I do feel like as a rule of thumb, taking less in the early stages. And then once you enter the growth phase, you accelerate for growth could be... something that I would do if I had a chance. Vignesh 2 (27:32) What was your example with like hitting that flywheel or that growth phase? How did you realize that? Mahesh 2 (27:39) It's basically you have to get to a product market fit point, right? Like what one of my seniors told me what product market fit is there are multiple ways to define it. But basically an average salesperson should be able to sell your product to an average customer on its own. So it cannot be founder led sales to a person who has a really strong pain point and things like that. has to be an average salesperson selling it to an average customer. if that happens, then I think you're ready to deploy a lot of capital to accelerate. Otherwise it could fall into the bucket of premature scaling. We see a of companies trying to prematurely scale and burn VC money. Vignesh 2 (28:31) The analogy that I'm kind of getting from this is, you know, the early stages is too early, but once you know that you have something, the incoming money is kind of adding fuel to the fire to like help with just the growth, not necessarily R &D or experimenting, but like you found the product, now it's growth. Mahesh 2 (28:50) Ideally with every round, you're trying to minimize some risk or the other, right? So once you get to a point where you're ready to scale, that's when capital could help you minimize, you know, risks. Yeah. Vignesh 2 (29:05) That makes sense. So what's next for you guys? What do you have in the roadmap for the next few years? Mahesh 2 (29:11) We are trying to get on the real world in a very real way. So it's not easy. So we have a lot of challenges in front of us that we need to solve. And I personally believe that if we execute on our product goals, there's going to be a lot of good things that come out of it. So I would say getting this to a real world deployment in a large way is the thing that is in front of us. Vignesh 2 (29:41) What are you guys kind of doing to solve that? Or what challenges are you facing with that? Mahesh 2 (29:41) And we'll We have gone into the real world and then observing where our product doesn't do as expected. And then we are seeing those issues. We are checking if it's a technology issue or if it's more like an integration issue. And then we prioritize that to see where we would get the maximum return for our effort. So that's basically what the plan has been. So get into the real world and... scale as expected. Vignesh 2 (30:17) And then do you have a vision for, you know, what does Vayu robotics look like in 10 years? it like every grocery store should have a delivery robot? Like what's the grand vision? Mahesh 2 (30:28) Yeah, I think, you know, like I said, dull. or dangerous tasks should be automated. And we are building the nervous system for it, where we are building the eyes and the brain for robots, which could automate tasks which are either dull or dangerous. And my vision is to penetrate into as many markets as possible using that core technology. And then as we evolve, go through this large language model phase, the new tools that get developed and to our advantage. So that's basically where we want to go. Awesome. Vignesh 2 (31:08) Well, thanks so much for taking the time to join today. If our listeners want to go somewhere and support you, check out your progress, what's the best place where they can find you? Mahesh 2 (31:19) So they could go to our website or our LinkedIn page. So those are two places where we keep updating. So please go to our page and then follow us and you get all the updates. Vignesh 2 (31:31) I'll add those links in the show notes here. Thanks so much for your time, Mahesh. Mahesh 2 (31:39) Thank Thanks a lot.

Since you scrolled this far, maybe you could leave a review?