Ali Kashani, CEO of Serve Robotics, on why sidewalk delivery robots are the future of delivery!

Okay, so we've got a new episode of Legends and Leaders and today is great to have someone here who's really pioneering an incredible new area of you know autonomous delivery. Having these

robots go around and um you know deliver stuff more affordably for people uh more efficiently and I mean they're now in so many different incredible cities. You

see them you know riding around and you're like wow this is the future. The

future is here. So I'm excited to have you here and again to all that you've built.

>> Thanks for having me. I love that the future is here is my favorite uh statement about robots.

>> Yes.

So just kind of like going back into your origins like were you always fascinated by robotics? Were you like a sci-fi kind of guy growing up? Like how

did you get interested in this space?

And then what made you want to turn into like something you actually pursued in life?

you know, um it it actually started for me when the first time I heard about u you know, Bill Gates and Microsoft starting a company out of his dad's garage and bringing computers to every

household. I remember my dad telling me

household. I remember my dad telling me this on a drive and uh I was living in Iran. So I was actually not in the media

Iran. So I was actually not in the media ecosystem that we have here. I have

never watched Star Wars, which a lot of people fault me for. But um but my imagination immediately went to okay, what is the next big thing? what is, you know, if the computers are in every

home, what is the the next thing that's going to have to be in every home? And

of course, that's robots. So, uh, that was that was something I wanted to do and start a company making robots and bring them to every home since I was basically a teenager living in Iran and,

uh, didn't know any better about any of those things of starting a company or making robots, but the dream kind of came to life, uh, at that point.

>> So, when did you come to the US? Like,

how old were you?

Well, I actually first moved to Canada when I was 19. And uh moved to the US, I don't know, 2016. So, been about 10 years.

>> And when what when did the company launch like for you during that time?

Like were you thinking about the idea, you know, 10 years ago, or was it something much more recent? You know,

I've been fascinated with robots ever since I um when I was in undergrad at UBC uh in in Canada, I actually started working on the DARPA challenge

competition. So, the self-driving cars,

competition. So, the self-driving cars, that's how a lot of uh really this this whole race began uh back in 2005, I think. So, 2004, five, that's that's the

think. So, 2004, five, that's that's the time frame. And I worked on a little

time frame. And I worked on a little known sensor at the time called LiDAR.

Uh which was this really industrial sensor from a German company and really expensive. And that's when it started to

expensive. And that's when it started to be adopted for self-driving cars. It it

became obvious that that it's uh it's kind of an unlock because you can uh understand the 3D geometry around you much better. Uh keeping in mind that at

much better. Uh keeping in mind that at the time AI wasn't very good. So looking

at a camera didn't really tell you much about the environment. You just saw a bunch of colors. Uh but with LAR you can actually see forms and you can understand if something is in front of

you. Um so I was working on this back in

you. Um so I was working on this back in 2005 basically or 2004 even um before anyone knew about that this this little known sensor called LAR. Um and then you

know I went on to start a few companies uh usually in the hardware space always in the AI space partly because I after graduation I decided that um uh you know

robotics wasn't quite ready uh for prime time and this is again we're going back to uh you know 2008 2009 when I actually 2006 when I finished my undergrad and

then 2011 when I finished my uh PhD um so I I didn't think the timing was right but then Um when I came to the US I uh met with the founders of Postmates they

were interested in my company at the time which was also in the AI space and uh and that's when this new idea was born which is okay robots on a sidewalk

delivering food it makes sense the time is right we should get going.

>> Were there certain like things technology-wise that like you felt like the time was right? Like what what really made you think the time was right? Was it just like the funding was

right? Was it just like the funding was there or like what what pushed you to say hey I now I can do this?

>> You know I think one of the biggest unlocks was the domain actually. It was

the fact that suddenly you had this uh incredible demand and growth in ondemand uh delivery. So you can you know anybody

uh delivery. So you can you know anybody can press a button and something would show up. But now it's being delivered in

show up. But now it's being delivered in a car. Uh as we like to say why move two

a car. Uh as we like to say why move two pounds in two time cars. Uh the fact is when you're moving things in a short distance of a couple of miles, you don't

need a two-tonon vehicle moving at I don't know 50 60 miles an hour. Once you

simplify the problem to a smaller form factor, less weight, less speed, you significantly decrease the risk. So the

the kinetic energy in a car is 3,000 times more than a robot of our size. And

kinetic energy is where all the risk is.

It's like hitting something. that's the

transfer of the kinetic energy that causes injury or possibly even kills people. But if you have thousands of

people. But if you have thousands of times less kinetic energy, you don't have that problem anymore, right? So, so

the technology that existed that wasn't quite there for for the cars to to be autonomous could actually work for the robots. And that was our thesis and I

robots. And that was our thesis and I think it's generally proven to be right.

>> Yeah. No, it makes sense. like just

building out the first robot like how did you finance that? How did you get that off the ground and did it really work well or was it just like kind of something that was proof of concept enough to get more funding?

>> Oh, it's been it's been eight and a half years. So, it started with really uh you

years. So, it started with really uh you know basic proof of concepts. Uh so,

Postmates is where where this began when I joined Postmates through that acquisition in 2017. Um we had some really interesting insights because we could look at the you know one of the

largest delivery platforms in the country look at the raw data of all the demand uh for deliveries and what do they look like and we could see that most of them are like a couple of miles

at most which means a robot can do it.

So that's that's where everything began and within a week like literally we started on a Monday. By Friday we had a robot uh on the sidewalk but it wasn't really very smart. It was being remotely

driven. Uh it was a bunch of Amazon

driven. Uh it was a bunch of Amazon parts that we had purchased. But the

idea was we should get something out so we can see what it's like to have a robot moving around on a sidewalk. And

we learned so much from moving so fast by by having something that we can actually experience the uh the end product very quickly and it informed a lot of decisions. So within I think nine

months we had done more than a thousand actual customer deliveries with our new fleet of little kind of prototype robots and uh you know couple years after that we had a robots operating in Los Angeles

during COVID and you know the rest is history. Now we have I should say we

history. Now we have I should say we have more than a thousand robots today by the end of the year 2,000 robots.

It's one of the largest autonomous fleets in uh any urban uh environment in the world.

>> How did you make it autonomous? Like

what went into getting that technology right? It's something the car industry

right? It's something the car industry still hasn't finalized yet.

>> Well, the a lot of the technology is very similar. It's you know you need you

very similar. It's you know you need you need to see the world. you need to then uh take that those sensor information and really understand the world and then you need to you know tell your robot

what to do uh and and and you know the the LAR sensor the better computers that that like Nvidia is is providing all that is is really very similar to what what you see in the cars but but the

difference is that we are not moving as fast and we can stop at any time and ask a human to help the robot. So, think

about it this way. If you're on a sidewalk or sorry, if you're on a street and you get to a intersection and you need to decide is is it safe to cross or

not based on what the traffic signal tells you, well, if you stop on a green, it's bad. And if you, you know, cross on

it's bad. And if you, you know, cross on a red, it's also bad. Um, and you really don't have a chance to just say, I'm going to stop right now and, you know, ask someone for help. That's not an

option because again, stopping on a green is dangerous. On a sidewalk, we don't have that problem. If we stop on a green, nothing bad happens because we're not blocking traffic. We're not moving

that fast. Uh crossing on a red would

that fast. Uh crossing on a red would not be great. So, we have models that are super sensitive >> to make sure we don't ever cross on a red. And that's much easier to build,

red. And that's much easier to build, orders of magnitude less complex to have a model that's really sensitive to one thing, but it can tolerate the fault in the other dimension. Um and our mo

basically our way of thinking our frame uh you know mental framework was when we are on a sidewalk we can have a human do a lot of work at first and our job is to

automate their work. So the very first delivery we did was fully teleaoperated someone was controlling the robot and our engineers would sit in that room and watch what that person was doing and

then decide what is the first thing we should automate that have that reduces the most amount of workload the fastest.

So that's the journey of kind of going from a robot that was purely human operated remotely to a robot that now is primarily you know doing the driving itself because we have slowly but you

know steadily gotten rid of all that workload for the remote uh operators.

>> Why go the route of partnerships with Door Dash with Uber um instead of you know I mean you could do building your own app and kind of be bringing people there. I mean, both could be fine, but I

there. I mean, both could be fine, but I mean, seems like the partnerships have been really crucial to the the business.

Like, why focus on that so much?

>> You know, I'm a bit of a data nerd, and it's something that I've always really, you know, enjoyed doing. On a weekend, you'll find me just, you know, playing with spreadsheets, even if it's not work. [laughter]

work. [laughter] So what what I what we did at the very beginning was we took the data from Postmates and we built some really cool

um like physical simulation models that would assume okay if you have 50 robots in this neighborhood and look at the demand from Postmates yesterday, what would have happened uh to these robots,

what would I what would they have basically done? And then we would, you

basically done? And then we would, you know, simulate different scenarios like, hey, if we have robots that are dedicated to specific restaurants or if

we have robots that are um, you know, that are, you know, deployed in a a certain way where in the middle of a day they have to stop and get recharged. So,

you know, that's that's a different robot design. So, there's a lot of

robot design. So, there's a lot of questions like how how much battery life should a robot have? How should it be deployed? uh should it work with one

deployed? uh should it work with one restaurant at a time per robot or should it work with all the rest? All those

questions we could actually play with the data and see what the impact of that to the unit economics looks like. And

one thing we learned for example is a robot should go all day on a single charge for the robot to make enough money. You don't want a robot sitting in

money. You don't want a robot sitting in the middle of a day for four hours or a person having to drive to a robot to, you know, swap batteries. That's too

expensive. Another thing we learned, it goes back to the question you just asked, which is if we were to have robots dedicated to restaurants one by one because we are going after one

restaurant at a time, >> it would not get enough utilization to make the economies work. And we could really easily simulate this like I remember I ran a scenario. What if we

get 20 of the top u chains in an area in in Los Angeles where we were operating and just assume we got them all under contract working with us. What would

that look like? Well, it turns out those 20 chains would have really 40 restaurants in the area and then I would put 50 robots. I'm like, they're barely getting the utilization I need. But if I

have those robots work for all the restaurants in the area, it's a night and day difference because there's actually a thousand restaurants there.

It's not just 40.

>> That's the key distinction. We realized

really early that if you are basically going and signing one restaurant at a time, your utilization isn't isn't going to work very well. But if you have a platform like Uber or Door Dash or at

the time Postmates giving you access to everybody, you would get the utilization you need to really jumpst start the platform and and have robots operating economically.

>> Yeah, the the approach makes a lot of sense. Um and like from the utilization

sense. Um and like from the utilization standpoint, it's just amazing like how the adoption from different cities has occurred. I think you've really done

occurred. I think you've really done great with going at major cities and getting them to be willing to do this.

How do you pitch them? Like what how do you convince them to to get on board? I

was just recently in Miami and I saw some over there too. Like is it just the economic benefit of it, the convenience of it, less maybe less um traffic in general? Like what what does it take to

general? Like what what does it take to get these cities to get on board?

>> Yeah. And look, it's it's actually um you have to be cognizant of a lot of the fears and concerns that people have.

some rightfully so, some I think are are more exaggerated like um you know people are talking about water usage in data centers and and there's some like really

crazy ideas out there um about what the reality uh you know really is. It's it's

it's really just not a big deal the way people have kind of made it to be. And

then there are other concerns around jobs or other things that are totally fair and we should be very mindful of.

So look, I think what's important when when you talk to cities or the public or just partners is to really explain the value proposition uh and help them understand why this thing needs to

exist. And it actually doesn't take too

exist. And it actually doesn't take too much effort to explain the the reasoning. The the line that I shared

reasoning. The the line that I shared earlier, you know, why move two pound burritos in two pound cars? It generally

is one of those kind of light bulb moments where people start to think, okay, yeah, this this is kind of crazy the way we are doing things. Because

think about it, it's a safety risk.

Every time you get in your car to go buy a gallon of milk, you are actually accepting the risk of an accident because that's, you know, that's that's

a reality of driving around and possibly even someone losing their lives. Like no

one who gets into a really bad accident, you know, knew when they got into a car.

Uh, and probably what they were going to do driving around wasn't important as as as you know, the risk of that accident.

And thankfully car accidents don't happen that often, but they do happen.

So our job is to get less addicted to these two-ton machines because ultimately they they're not really helping us move the move the gallon of milk. They're moving us safely and

milk. They're moving us safely and comfortably, right? So when you explain

comfortably, right? So when you explain that to folks, I think it it it really rings a bell. And then of course on top of that is the fact that you can reduce congestion. Everybody wants to, you

congestion. Everybody wants to, you know, drive in streets that are less congested. It reduces emissions. uh you

congested. It reduces emissions. uh you

know we want to have greener cities and it has an economic value for the restaurants for the mer for the for the customers and by the way also for the workers in those cities because we're

going to create much higher quality jobs than uh you know a a career today moving one burrito at a time.

>> Yeah. So there's a lot of benefits for for cities very clearly from what you just said. Um so I think yeah I mean

just said. Um so I think yeah I mean you're just going to keep scaling it to more and more cities as as you go. When

I was recently in Miami, uh there was a a serve robot that needed help crossing the light. Um it was the light was like

the light. Um it was the light was like it was just crossing. It needed like someone to click the basically the walk button. I thought that was really

button. I thought that was really fascinating cuz it asked me to do it and like shows you on the screen. I was

like, "Wow, this is really interesting the way we're interacting with robots now." Like how do you think about robot

now." Like how do you think about robot human interactions and like stuff like that? Like why even have a feature like

that? Like why even have a feature like that?

>> Yeah. First of all, thank you for your service [laughter] pressing the button. Um, look, there there are c certain things like, you know, the the manual intersection

buttons. We are working on on automated

buttons. We are working on on automated solutions for those as well. But, uh, I do think that, you know, when when you're operating in in public environments, there's always going to be

moments where, um, there's an opportunity for for some collaboration.

And maybe uh today it's it's you know people helping robots and in future hopefully the robots also helping people. There there are all sorts of

people. There there are all sorts of scenarios that that we could be valuable for and besides you know moving goods around. We do that today by the way we

around. We do that today by the way we share uh data with cities about uh you know infrastructure issues or where there is no care cuts which is valuable for people on wheelchair or or with

disabilities. So you know we we're

disabilities. So you know we we're trying to make sure that we are also a good citizen helping other people. uh

but at the end of the day we've designed a robot to be easy to accept uh you know by society. In fact, I think it's a

by society. In fact, I think it's a missed opportunity if we don't. It's our

we our imagination has been so ahead of us for so many years with robots and for us to kind of make a really boring machine when we can have a little bit more fun and make it something really

exciting. And I mean just imagine a

exciting. And I mean just imagine a four-year-old child seeing a robot for the first time. I've seen videos of this so many times with our robots where a child kind of runs in front of a robot and is so curious and excited and I mean

that's one of those experiences they're never going to forget. And what a miss not to really lean into this and make the robots fun and cute and and and then as a result, you know, something that society has an easier time accepting

because again we talked about the value proposition. I really believe that these

proposition. I really believe that these are important for cities to to kind of again make cities better, make them more pedestrian friendly, make them safer for everybody and of course, you know, give

that benefits to to the to the restaurants and local merchants. So, but

we have to make something that people want to adopt. And so, the robots have eyes, they have names, they're cute, they're polite. All of that is is part

they're polite. All of that is is part of it. And by the way, you helping a

of it. And by the way, you helping a robot probably also helps you maybe fear the robot less because you know, you know that this is this is something that that kind of collaborates with you.

>> Yeah, I thought it was quite nice actually that it asked me to help it. I

don't know. I just it it felt interesting and I I did it, you know, there was nothing to lose from it. I was

like, "All right, I'll click the button, you know." [laughter]

you know." [laughter] Yeah. Um physical AI is an area that I I

Yeah. Um physical AI is an area that I I personally like a lot. It's an area that you're really in. Um I think it gets overlooked a lot physical AI. I mean

there's so much happening in the software space. It's easier I guess to

software space. It's easier I guess to scale to millions of people in the software world. You know why do you like

software world. You know why do you like physical AI and building in the physical eye space?

>> I mean because B2B SAS is just [laughter] you know it's only going to take you so far at the end of the day right at the end. I think the every time we

have a new um kind of paradigm shift in in in computing whether it was first was just creating you know the the processing uh units and therefore the

the computer uh so we can you know do math faster or run a spreadsheet and then we connected everything together that was a big step and then we made them smaller so it fits in our pocket and now we are making them smarter so

every one of these paradigm shifts I think it starts with impacting our digital lives first, but eventually it's going to impact our physical lives. And

I think that's where the biggest opportunity ultimately lies. So it is much harder. Absolutely. But but it's

much harder. Absolutely. But but it's also a lot more fun. And I feel like the the impact of it is just so meaningful.

Like think about how cities changed because of cars. This is that was like a physical technology that came in the engine really kind of started all that.

We're looking at that kind of magnitude change. And I feel like I'm so fortunate

change. And I feel like I'm so fortunate to be sitting at this moment of time working on something that is going to have that kind of impact on society and how we live our lives and how our cities

look and feel. So uh I don't know it's hard to not get excited about about physical kind of um AI or physical uh anything like robotics.

>> Yeah, I agree. I like to make stuff and then you know it's it's there in this reality like it's something real. You

can touch it. You can feel it. It's it's

quite exciting. Um and like you mentioned the disruption opportunities are a lot especially I think there's less competition too in general because it's just the barrier is higher.

>> Um when you think about like building like just just like scaling this in general like do you see different types of robots being needed for different applications because I know you have

like one primary one now that's used on sidewalks. Um, have you kind of thought

sidewalks. Um, have you kind of thought a lot about that or is there not really much of a need because the current version serves pretty much most areas?

>> No, I mean this is just first the first form factor basically and and you know every generation of the robot is going to be more capable but there's then going to be all sorts of you know more customized or more generalized actually

you know you may you may have a you may want a robot that can do more things uh when when the frequency of that work is is less than you know uh the initial

applications. I think um there's going

applications. I think um there's going to be robots in every human environment that that we we we have right like you have your maybe vacuum cleaner today at

home which even that you know has has a lot of limitations today and there's a lot of room for innovation but there's so much more that we could have robots

assist us with and we partly are blocked by the fact that robots have a hard time navigating our 3D environment and and coexisting with us in a way that's not

disruptive to us and and that and they're actually successfully and reliably able to accomplish their mission. So, uh that's the problem

mission. So, uh that's the problem ultimately we're solving because if you get robots moving around sidewalks in in major cities, I think you've solved one of the most complex environments and

bringing those robots to other places uh even if their shape and size and and kind of physical capabilities changes the brain behind it is largely the same.

I think that that opens up a lot of a long tale of use cases uh that that I'm very excited to to see this platform kind of adopted to.

>> Yeah. Yeah. It's it's quite exciting. Um

just thinking about like the the scale of this and and how much you've already scaled so far. Um has being a public company been helpful for that at all? Is

it is it like why did you want to you know be a public company in general? And

what do you think you even need I mean to get to this next phase of scale? Is

it a capital problem or is it just a city cities being okay with it? Like

what is the bottleneck?

>> Yeah. You know um so this is my maybe fourth company uh that that that I've helped start. So one thing you realize

helped start. So one thing you realize is the innovation cycle and or or life cycle and and the capital life cycle are not always a match. So investors in the

private market, they get excited about something for a couple of years, then they move on. And your job as a founder is to last something like 10 years before you get to a place where you're off and on your own feet. So how are you

going to bridge that gap? And different

people find different solutions. I

wouldn't say every solution is the same.

If you're very fortunate, maybe you can do that in the capital in the private capital markets and and and you know get yourself to the destination. But um but sometimes you have to to look outside

and see you know how to solve that mismatch in the in the life cycles of private money versus uh you know uh startups. Uh in our case we went public

startups. Uh in our case we went public in the worst year to go public basically in terms of the number of IPOs but it was also a really uh you know difficult year in the private side with um I know

many companies in our space in the autonomy space that started roughly the same year as 2017 and they didn't make it. Some of them were inside bigger

it. Some of them were inside bigger companies like Amazon had an effort with sidewalk robots for example. Google had

a robotics team that they shut down. So

it didn't matter if you were part of a big company or you were an independent private company. It was a tough market.

private company. It was a tough market.

So our thinking was uh we should try to access the largest pool of capital because some people are going to understand this and others don't. And we

don't need everybody to believe in this.

We just need those few who really get the vision. And when you're in the

the vision. And when you're in the public markets, you get access to all investors rather than just the private investors. So, it was a bet. It took us

investors. So, it was a bet. It took us a year to kind of go through the process and we wouldn't have known until the end of the process whether the bet paid off.

Obviously, it really paid off in this case. I I'm not sure if it always would,

case. I I'm not sure if it always would, but you know, we got fortunate and and we got to that result. But I would tell you one more thing. I always wondered about this be long before we decided to

go public which is we're going to new cities and you know there are merchants there are people who are going to interact with our robots how do we make

sure or try to help them you know have part of the upside of what we're building here other than obviously the economic value to those merchants and

customers. Um and so this idea of being

customers. Um and so this idea of being a public company so that more of you know the communities that we are going to be part of have an opportunity to

join this ride and join this vision was very compelling from basically when we first became an independent company by spinning out of Uber uh back in 2021. So

it was something I talked to our you know head of comms almost uh as soon as she joined I'm like look I think we need to go there but I don't know how it sounds kind of crazy but it's something that is in the back of my mind. So when

the opportunity came for going public it was already something that we were we were thinking is a good solution to solving the the broader problem of you know going to communities and benefiting

people you know in in a in a more meaningful way.

>> Yeah. Well I appreciate you taking the time and doing this. That was that was all the questions that I had. I think

it's amazing what you've been able to build here and how you've created this autonomous network that is I think the most advanced real world use case autonomous system fully autonomous

system in many cases that there is. Um

and it's it's changing the way delivery works in general. It's having that type of effect that you mentioned before and I'm just excited to see it continue to scale and thousands and thousands of more of these robots out there and

what's to come next. Thanks for coming on. Thank you so much for having

on. Thank you so much for having