325. The Robotics Founder Who Exited His Last Company to Twitter w/ Zehan Wang | Paddington Robotics

0:00I'm Zehan Wang. I am one of the founders of Patent and Robotics and we're currently working on solving robots for retail and supermarkets with the grand goals of actually solving robots in general human-centric environments. And like how's it been going so far? It's been a very interesting journey so far. First of all, figuring out what we actually wanted to do in robotics. It's quite a vast space. And when you think about the opportunities for robots in the physical world, and the physical world in general, there's way more businesses and opportunities in this world than just a pure software world.

0:36And so part of that was figuring out where we would even start, what is the scope of what a robot should be doing, and how does that allow us to really scale up and generalize to, you know, say you want a robot butler in your house in the future someday, like how does solving it in a supermarket allow you to do that? And I think part of this is figuring out like, you know, the bits and pieces, how we want to tackle this more systematically. And ultimately for myself, like, I don't really believe humanoids today are the answer that we're looking for in, you know, 20 years time, that we're a little bit premature, like we're not going to get there.

1:10And by the time we get there 10, 20 years time, the form factor of these humanoids, or even if they are humanoid, will be completely different. Tell us a bit more about that. What do you mean the form factor? So I guess a lot of people think robot, they think like movie robots, right? Yes.

1:25Amardeep Parmar:But then you're saying that they should be in different shapes. So tell us more. Yeah. So I think a large part of form factor needs to be really driven by the use cases that you have in mind. And it's almost sort of very lazy thinking to say, OK, we're humans and we do a lot of things. So we should build robots that look like us because they could do a lot of things. But then when you really think about it, humans aren't actually very good at doing many things physically. We're worse than many animals in terms of strength, in terms of dexterity, and in terms of agility. And if we build robots like us, they're also going to be somewhat average in terms of what they can do.

2:02And the saying is, you really want to have the right tool for the right job. and that's the same whether it's a human doing it or whether it's a robot doing it and so i think in the future as we you know part part of the journey that we're on is actually figuring out what are even the right use cases where does it commercially make sense and just going okay we're going to build a humanoid because it looks cool you obviously get you know a bit of a hype you get sort of that novelty factor but then if it actually doesn't perform commercially you're just going to be stuck with a lot of you know humanoids it's sort of like you know when we had loads of scooters and they all just ended up in canals.

2:39That might be what might happen to all these humanoids eventually.

2:42Amardeep Parmar:You said about what's the best use case to start off with, right? And it's also about, well, I guess one is like, where can you get the money from to actually build it from? And how do you go about thinking about that? Like, okay, because one of the classes in the robotics, obviously, it needs a lot of funding up front. And how do you think about what way to approach that? Did you go, let's try and raise as much money as we can? Or do you think like, let's try and get early contracts? Or how did that work for you? Initially, we're thinking, well, let's just try and raise a lot of money. But then the thing we realized is VCs have no idea what's going on in robotics.

3:12And in some ways, they're just very confused. So at least when we were initially starting, there weren't that many VCs who had high conviction about what direction robotics should go in. They're kind of all just, let's wait and see. Show me some traction. Or do something else that's cool. It's not enough to just have an idea. and part of it is also driven by the fact there's been a large influx of new robotics companies in the last few years. So for VCs, it's very difficult to understand what's going to be a winner, what's not, what's just people dabbling, because now it's easier than ever to start dabbling with hardware and robotics.

3:55But yeah, I sympathize with them.

3:58Amardeep Parmar:It is difficult to choose the right winners in the space. Obviously your background means that people might think you're more likely to win another. And can you tell us a bit about Magic Pony and what happened there and why that kind of sets you up as well? Yeah so 10 years ago in fact I exited my company Magic Pony which worked on super resolution and video compression. We were one of the first teams to really be building methods using deep learning for doing you know this type of image and video processing what we're doing we were doing back then would now be called generative AI but we didn't even have that term for it back then we basically you know when less than a year between raising our seed round to get acquired by Twitter and we exit the company for actually I'm technically I'm not meant to say but if you look up online you can find the numbers I was like am I allowed to say that I've seen what I've seen on TechCrunch.

4:51Amardeep Parmar:Yeah. So I've seen over 100 million on TechCrunch. Yeah. So, yes. So obviously having that kind of an exit, right? There's not a huge number of entrepreneurs in the UK with those kind of exit numbers. And with VCs and obviously have the classic thing of where so much, well, as you know, many VCs don't actually return any money to their whole piece at the end of the day, right? They might have money, which they got on paper, but they didn't actually get an exit. And you're one of the few people in a way who've got that kind of experience. And when it goes to those discussions with VCs, and we talked earlier how you just saying you're going to build something obviously will attract a lot of attention.

5:32Amardeep Parmar:And how has fundraising been this time compared to those first times you did it? Yeah, I mean, I think when we were fundraising for Magic Paint, this was like 2015. There There definitely were fewer VCs in the UK and Europe. And we certainly would have found it difficult to attract attention of American VCs as well. And then even in terms of angel investors, there were far fewer back then too. So there was a very limited selection. And we really limited how much we were asking for. But eventually, we built that number up. So when we first got term sheet, it was like 300K. We're like, yes. And then we got another$500 ,000.

6:13We're like, OK, doing better. Then we got some million. Like, oh, OK, we now might actually be able to do something with that money. Particularly if you're hiring, well, back then hiring good software engineers, scientists, et cetera, definitely was cheaper, but still not cheap. And now it's like even more, way more expensive to hire good people. And I think one of the things that's definitely different this time around is, yeah, way more investors out there but then there's also a lot more investors who are new at the game they're still figuring it out um the more seasoned ones maybe a bit more wary of making you know uh taking leaps but a lot of them are restricted by the rules that they set with their lps as well so yeah so it varies across the board um but you don't see as many like you do in the us who are just like oh yeah we'll just throw 20 million and you know it's a safe and there's like nothing behind it at all.

7:08I think in the UK, yeah, that's pretty, you know, very, very rare to see anything like that.

7:13Amardeep Parmar:And with that magic pony too, you were almost the golden like poster person for the entrepreneurs first program as well. The kind of proving their model out, right? Like obviously it's been very different this time in terms of having a co-founder compared to the first time around. How are those two experiences and like, what can you take from your first option to help your second time around? As a first time founder going through entrepreneur first, but you know, It made total sense for me. I had zero idea what we're doing. And in fact, neither did my co-founder at the time as well. So pretty much everybody at that time just figuring out, even for EF at that time, they were still figuring out.

7:46It was like a startup running other startups and trying to teach startups how to be successful before they were successful. So it was definitely an interesting time point. But there was a lot of thoughtfulness and thinking behind things to try and what to do. We were, I guess, one of the first deep tech companies to go through EF, which doesn't actually fit well with the EF program at all. You have 12 weeks to come up with an idea and decide whether you're going to go through to demo day or not, whereas most deep tech companies take far longer to really get the ideas to a point of maturity that makes sense to then go for more further investment or whatever else you might want to do.

8:27So, yeah, we were kind of the exception to the rule for EF back then, but then we managed to make it work. It's definitely very different times now. EF still have this 12-week program, but they're now looking to bring... They want some of that deep tech caliber talent, but without necessarily like, oh, you're going to start a whole new research project from the beginning when we joined EF, but rather to continue those experiences and look at how you can apply it more immediately. Right.

8:55Amardeep Parmar:And then co-founder for this new company, how do you guys meet? There's a few of us who are co-founders. So initially I just asked Johannes, who was one of our first engineers at Magic Penny, whether he wants to join me for another adventure. He was with us from the very early days of Magic Penny through Twitter. He stayed on Twitter for actually longer than I did, I think six years. And then he went to InstaDeep after Twitter. And then after that, we were like, okay, let's do something. We actually decided over a beer that we should get into robotics. And then the thing we realized, well, neither of us have a hardware background.

9:34We should probably find someone with hardware expertise to come join us. Oh, that's Jenny as well. And then I reached out to actually someone else who had just been through the EF program but didn't find the right team. And then also someone who had a previous robotics company but didn't manage to make it a success, partly because they were too early. And also because they struggled to raise funding for what they were doing back then. So now there's like four of us who are active co-founders and then we've now hired a team and there's like about 15 of us.

10:09Amardeep Parmar:And how are you enjoying it this time around with the new co-founder dynamic and coming back into the game again? Yeah, it's definitely different. Like it's not a sort of equal splitting co-founder. like it was with Magic Penny between Rob and I. But that's also partly because I'm putting more of my own capital behind what we're doing to get us started. I'm in sort of a privileged position where I can do that, whereas before that was obviously not an option. I'm able to really sort of think about this in a different way than how I would have done at Magic Penny. Like, I don't have to accept the first VC offer I get.

10:46I can actually say, you know what, I could just start funding this myself. I don't have to wait till, you know, what we did with Manchapini was like we spent, you know, three months on the road Pitching to VCs and eventually nothing was being done the whole time we're pitching All right, no work was being done. We had to go with these demos that we just you know, some MATLAB demos that we had But you know, Eventually it worked, but that was like say, you know, its own set of stressful problems. Whereas now I'm just like, well Let's not wait. Let's just start building and I can just fund us to get us going and we'll figure out as we go along.

11:22But then we can move much faster and don't have this weight between fundraising and actually making progress.

11:27Amardeep Parmar:When you are in that position that you are in now as well, what makes you want to go and raise VC again as opposed to other ways you can do it? Part of it is scale and ambition. It is where I want to go is going to be beyond what I can really fund myself. And I think the other side is getting that feedback, I guess, from the community that I'm not just a crazy person. And that's actually, we are going to be doing something that people can believe in, that people are willing to back. And so it's always good to have external help. I guess one of the things you learn as a founder is like most success stories aren't just the founder plugging away.

12:11They always have a network behind them that are able to leverage that helps them along in that journey, even a tiny little bit. But there was always that, and investors do play their part in that network.

12:21Amardeep Parmar:And I guess when you're thinking about hiring the team that you hired too now, this time around, what did you look for this time because of the mistakes you made maybe last time? You know, when it comes to hiring people, we're always looking for the best in terms of what we're hiring. And, yeah, there's always a chance, you know, you have a limited window in order to assess somebody, whether the right fit. Sometimes it's not the fault of the individual, just, you know, the stage of the company that you're at or what it is you're trying to do. and even you know it doesn't matter how enthusiastic they are sometimes that just doesn't quite work right and part of that doesn't really play out until you spend you know sufficient time with them and see that we you try to do as much as you can you know during the interview process and and obviously we try to maintain very high bar on in terms of that hiring standard but yeah it's uh it's there's never 100 % foolproof what like do you think you index on more than maybe other people do when you're hiring it's hard to know what other people do to What really matters to you, Agnes?

13:18But more generally, I would say for us, because we really focus on developing technology, but not just let's take something off the shelf and make sure that works, but also how do we carry this forward? How do we innovate on top of that? So that deep tech expertise, if you will, that experience with technology is a key part. But on the flip side, we need people who can actually get shit done. We don't need theoretical academics who are like, oh yes, this is the best method, but I don't know how to implement it. We actually need people who can get their hands dirty and not be afraid to do so. So there's a bit of a balance.

13:59Sometimes at times you feel like you're looking for a unicorn, and you can't just, it's obviously hard to find. So you kind of have to build a team around strengths of who is already on the team, what you might be lacking. and in terms of that team building, you want to really cover for each other's weaknesses. And that's really how I really think about hiring. It's not just an individual isolated in themselves, but how do they fit into the wider team? Are they a complement to the team? Do they bring something new to the team? Or are they a slightly different version of John who may not add that much, right?

14:38Amardeep Parmar:And with your own role now as well as you're hiring out, what's the stuff that you're focusing on like on your day-to-day like where's your attention going it's yeah as as always an early stage company it's all over the place you know where right now like we're looking at new offices i'm preparing for fundraising we're still got ongoing customer discussions we have a pipeline of you know customers as well so part of it is you know finding a balance but we're so you know a small team now that not everything is just on me so and part of it is to figure out delegation, how each one of us can sort of take on a little bit and then really, how do we come together as a team?

15:19So there's still a lot in terms of team leadership and making sure everyone is able to come together, people feel empowered, people feel like they are getting the right risk reward for what they're doing as well. Part of that is just making sure we're making progress at the end of the day and it's going in the right direction.

15:39Amardeep Parmar:Hello, hello. I hope you've enjoyed the show so far. I'm Amadeek Parma, co-founder of BayHQ and the host of this show. For those of you who don't know, BayHQ is the community for high growth Asian heritage founders and investors in the UK. Over 7 ,500 people have attended our events, 200 people have been for our impact programs, and obviously there's been over 250 episodes of this podcast. If you want to join us and take part in the program and come to the events, go to www.bahiehq.com forward slash join. We also just released our first ever book called Startups for Outsiders, which you can get on Amazon now and the link is in the bio.

16:20Amardeep Parmar:Hope to see you soon at our event. Hope you enjoy the rest of the episode. Also with that as well, so with their selling side, right? So you said it was a different selling process to Magic Pony for you now and you have to learn new skills. So how does that look? In terms of the customer profiles working with, it's very different to the type we're dealing with MagiPony. So during MagiPony, we actually initially tried to work with telcos because we thought they felt the pain point in terms of increase in video footage online. I think even back in 2014, 2015, I think 80 % of all internet traffic in North America was actually Netflix data.

17:03Now, with not just Netflix, YouTube, social media, etc., way more internet traffic is video or rich media data. But then we also realized that telcos felt the pain, but there was actually very little they could do about it. They just had to put up and provide the infrastructure. But then it was at the application layer where they could actually do some tweaks to performance, look at how to stream that content and so on, and then look to provide quality advantages versus the competitors. So we had to do a little bit of pivot with management in terms of who our customers really were, what is the long-term product goal for us.

17:45Whereas now with Paddington, we decided very early on we wanted to target retail. And one of the reasons for that is we're looking at retail, the shop themselves, not the backroom or the warehousing, but really looking at how do we solve this problem of having robots that can work around people and work in human-centric environments. So that kind of sets the customer profile that we have to go after. So part of that is making that balance, making sure we understand how to deal with these customers. customers and a lot of this is enterprise sales which requires more relationship building and sort of you know finding the right partners to sort of build out these relationships with.

18:24Amardeep Parmar:Because I imagine that when you go to like you can't just pop into your local Tesco and be like hey like can we put a robot in here right? Yeah. And like you have to work for the different layers and I guess your experience like existing experience and the VCs and things you know has that helped at all has been kind of cold getting into this space? Yeah it was actually pretty cold. We made a first contact because I just found some person on LinkedIn and I just reached out. And then initially we just had to do cold outreach via LinkedIn and try to look up the right profiles of who we wanted to hit.

18:55And then as with cold outreach, you might eventually get someone who takes the bait and then you're like, okay, let's have a chat. Let's sort of figure out how it goes forward and so on. So that's sort of how we got started. Yeah, you definitely can't just take a robot to a store and just say, hey, can we do some trials in your store? Yeah, I'm not sure they'll allow you to do that. But who knows? Maybe some store managers are like, yeah, yeah, go free or whatever. I feel like there's kind of some viral TikTok series here. Yeah.

19:23Amardeep Parmar:Take the robot into the store and see what happens. And then, yeah, I'm sure there's something like from a health and safety perspective like that shouldn't be allowed. I mean, one thing we're learning now as we go through, you know, these trials in store, there's a whole bunch of things that are ticking off from the health and safety, the property, the fire, the insurance. So because you're taking, it's a robot, but it has a battery, which could catch fire. It has motors. You have to think about, is it likely to be a risk for other customers? What about property damage, et cetera? So yeah, it's a bit more complicated, I would say.

20:04That's one thing we thought initially, like, maybe we could just rock up to a store with a robot. And then we realized, oh, no, yeah, we should probably do things by the book.

20:13Amardeep Parmar:You said as well how hardware is often considered a lot more difficult than software, right? Yeah. And I guess you obviously knew that before you went to do Paddington. How is it the reality compared to what you expected been there? I guess in some ways I thought I needed more of a challenge this time. So I thought, let's go try to do something with hardware. So, yeah, it definitely has been more challenging in a way. And part of it is also getting used to sort of how hardware development is done, but also dealing with, you know, when things go wrong, it's not just, oh, you fix a line of code and that'll be done.

20:45There's more to sort of from a debugging standpoint of like, oh, actually one of the cables came loose or something like this, right? Or like we actually damaged, you know, one of the components during transport or something like this. Yeah, there's a number of things now you have to be on the lookout for that you previously, as someone who worked more in software, or even just in the data and the modeling side of things, you didn't necessarily care about as much. But in other ways, this is kind of good because we're now experiencing what the real world is actually like. This is embodied AI in action.

21:19you know and these are all the sort of pain points that robots will eventually have to solve in order to become you know widespread in wider society and looking at the wider society angle right with