Startup Explore

Collin Barnwell (00:00)
So we make AI powered traffic lights. We are trying to reduce travel times, improve efficiency at intersections, and improve safety.

we're building a 3D real-time map of everything that's happening at the intersection. So every car, bike, bus, pedestrian, as it's moving around.

and then we're using that to basically run a simulation 10 times a second to figure out what the best light change is to optimize for safety and travel times.

And so what we see is a higher percentage of people running red lights, like right in that time where it's like, should I stop? Should I go? there's a high risk of somebody running a red light, which causes issues with the conflict zone. And so we try to just avoid that situation.

our first deployment went live in San Anselmo, California

our second intersection is going live tomorrow

Ali | Startup Explore (00:49)
Meet Colin Barnwall, the CEO and co-founder of Roundabout Technologies. It's a YC-backed seed stage startup that is using AI and computer vision to optimize how traffic lights work and make streets safer for pedestrians, drivers, and cyclists. Let's begin.

Ali | Startup Explore (01:06)
thank you so much for coming to the show.

Collin Barnwell (01:08)
Yeah, absolutely glad to be here.

Ali | Startup Explore (01:11)
All right, let's dive in. So the first question is, can you tell me what you're building and what is that big problem that you're trying to solve?

Collin Barnwell (01:19)
So we make AI powered traffic lights. We are trying to reduce travel times, improve efficiency at intersections, and improve safety.

Ali | Startup Explore (01:28)
can you explain how that works? What's the technology behind it? And I guess what kind of hardware you would need to have in place.

Collin Barnwell (01:32)
Yeah.

Yeah. So existing intersections will often have detection systems. Those detection systems detect the presence or absence of a vehicle in a fixed location. So sometimes this will be an inductive loop, a piece of wire underneath the road. says, yes, no, there's a car here. Sometimes it's a camera and you draw a box and you say, yes, no, there's a car here. And then, ⁓ the traffic engineer uses a combination between that information, which is like binary simple information.

and historical traffic studies to figure out the best way to time the light. ⁓

As you can imagine, that data that you're getting, this yes, no, there's a car here, is a pretty limited subset of, like it's not quite enough for you to make a good decision. ⁓ And so what we try to do is we use, we also use cameras, but instead of detecting the presence or absence of a vehicle, we're building a 3D real-time map of everything that's happening at the intersection. So every car, bike, bus, pedestrian, as it's moving around.

where we're seeing it, the whole field of view of the camera, we're mapping it into 3D space, and then we're using that to basically run a simulation 10 times a second to figure out what the best light change is to optimize for safety and travel times.

So ⁓ every traffic intersection has a refrigerator-sized cabinet next to it that has a thing called a traffic controller.

⁓ which is the thing that gets programmed by the traffic engineers, has, ⁓ and it has the detection inputs. So we essentially replace those detection inputs with our system, which is a GPU compute unit that is taking in camera data, building this 3D model, deciding when to send detections to that controller.

Ali | Startup Explore (03:25)
Got it. So GPU compute unit, that's a hardware piece of roundabout. I see. Nice. Now, if I double click on some of the features that I saw on your website, ⁓ specifically two of them, the dilemma zone protection, and then the dynamic all red extensions, can you explain how those two work?

Collin Barnwell (03:28)
Exactly.

Yeah.

Yeah. So those are two of our safety features. ⁓ Dynamic all reds are so a huge, a huge percentage of injuries and deaths happen at intersections and signalized intersections in particular.

A lot of this happened when people going different directions are in the conflict zone at the same time. It's called the conflict zone because there's conflicting directions of travel. ⁓ We try to get rid of that by one of the things is dynamic all-red extension.

So this is basically...

If you have somebody that's getting into the intersection relatively late, either because they are moving slower, like maybe they're a bike, right, and there's a three second yellow light, and then there's a one second all red in every direction, and then the cross traffic turns green, but maybe it takes the bike seven seconds to get across the intersection. And so that cross traffic turns green when the bike is still in the intersection. ⁓

That's a situation where that one second all red, so red in every direction, is not long enough to keep that person safe. So we would extend that ⁓ until that person can either get out of the intersection or until our maximum already extension elapses. So that's a situation that can happen not just for bikes, but also for...

trucks that are really slow to make a big wide turn or even just cars that go a little bit late into the intersection. So that's all red extensions. The dilemma zone protection is,

Ali | Startup Explore (05:06)
Mm-hmm.

Collin Barnwell (05:18)
There's two dilemma zone types. One is type one. That is basically, it is,

There's a certain yellow and red light, ⁓ you know, timing where you actually could conceivably get a yellow and a red such that you don't have time to stop before the red and you will run the red

So those are bad and they should never happen. They can happen, but intersections are designed to not have that. So I just want to be clear because ⁓ we don't like operate in that kind of world where that's a possibility. We operate in the type two dilemma zone, which is where similar situation, there's a yellow and then a red light and it's sort of close. It's sort of hard to decide if you should stop or go. ⁓

And so what we see is a higher percentage of people running red lights, like right in that time where it's like, should I stop? Should I go? Some of those people say I should stop. Some of those people say we should go. And that's a, that's a time where it's, there's a high risk of somebody coming to a sudden stop and getting rear ended, or there's a high risk of somebody running a red light, which causes issues with the conflict zone. And so we try to just avoid that situation. We map and project with our 3d projections, where's somebody going to be.

a couple seconds from now and we basically move trips out of that zone so nobody's like that close to having to make this decision.

Ali | Startup Explore (06:48)
I see that's pretty smart. So the all red extension is basically you try to keep red for all of the directions until someone or something clears the intersection. And then the, ⁓ the zone protection dilemma zone protection is basically where you try to, I guess, take into account the, direction of the traffic, the speed of the traffic and those seconds that might need it for ⁓ traffic light to.

Collin Barnwell (06:57)
Mm-hmm.

Mm-hmm.

Ali | Startup Explore (07:17)
become red from yellow.

Collin Barnwell (07:18)
Right, we're projecting

that somebody, they keep going at their current speed, ⁓ they're gonna miss that yellow light by a little bit, run the red light, and be in the intersection. So it's just kind of a bad situation for whether they stop or go. Yeah.

Ali | Startup Explore (07:23)
Mm-hmm.

That's pretty impactful. feel like those few seconds are pretty, pretty significant when you're driving a vehicle at certain like, especially at the higher speed, I guess. Okay, let me throw a couple what if scenarios. And the first one is if it's a bad weather, the visibility is low, it's foggy, for example. What does that mean for roundabout technology and systems? How does that impact?

Collin Barnwell (07:43)
Mm-hmm. Yeah.

Yeah, so we're pretty robust to rain, snow, fog. Basically we can see as well as a human can see. ⁓ But there are situations where a human can't see that well because the fog is thick, you know. ⁓ I should say we probably see a lot better than humans do at night.

Ali | Startup Explore (08:19)
Okay.

Collin Barnwell (08:20)
For fog, that is a good point because you can have thick fog where maybe you can't even see 50 feet out. And in that case, we... ⁓

Ali | Startup Explore (08:23)
Mm-hmm.

Collin Barnwell (08:32)
we basically detect that we can't see anything and we go into what's called a maxed recall, which is where we're just sending what's helping controller. Look, there's people waiting here ⁓ to make sure that every direction still gets a green, even if we can't see those people. And ⁓ that is no longer going to be optimal for.

travel times, but our thinking is sort of, well, if you can't see more than 20 feet in the future, not optimizing traffic flow for throughput is maybe not the worst thing in the world. And the important thing is that it's still safe and everybody still eventually gets a green light.

Ali | Startup Explore (09:12)
Got it. So if I say in those kind of situations, you kind of default to the existing traffic light system, would that be pretty accurate? You kind of stop trying to optimize for the throughput and safety as much as you would if those conditions were not there and you kind of default back to like the default position of traffic lights. Got it.

Collin Barnwell (09:27)
Ahem.

Yeah, that's fair.

Whatever the timing plan says,

Ali | Startup Explore (09:37)
Yeah.

Okay. And a few other scenarios. What about the emergency vehicles? Let's say ambulance coming with a siren on and how do you guys detect that? And what's the action there?

Collin Barnwell (09:50)
Yeah,

so that's a great question. ⁓ we are working on implementing the ability to actually detect with vision an emergency vehicle with its sirens on. We haven't released that feature at any of our intersections yet, ⁓ but we are looking forward to supporting that. ⁓ There is...

So other systems can support what's called like emergency vehicle preemption, which is they will get a radio call or a call over ⁓ some kind of a computer network to say, hey, police cars coming, like give this direction or red or whatever. And those calls come in as higher priority than our detection calls. So if a city already has that system, those calls are going to override whatever we're telling the controller.

But yeah, in the medium term, we would like to support actually being able to see those vehicles and make those calls ourselves.

Ali | Startup Explore (10:51)
Got it. Now, ⁓

what about the cars that are coming at a higher speed than than average traffic? How do you handle those cars and how far can you actually see those cars coming at a like higher speed?

Collin Barnwell (11:05)
So we can see like I said our we can see about as far as a human could see so our Distance tends to be limited by geometry like where's the next curve in the road or the next tree that's overhanging everything? ⁓ But at our current deployments, it's several hundred feet in every direction And we can see those vehicles from you know that far away that winds up being

Ali | Startup Explore (11:16)
Mm-hmm.

Collin Barnwell (11:34)
five to 10 seconds away, probably closer to 10. ⁓ we, because we, because we can see the position and velocity of everything, we can support, and this is something that we've talked about with, some cities, the ability to, ⁓ you know, respond to that in some way, which could be dynamically, you know, throttling the speeds of traffic by, by giving that direction a red light. If there's time to.

cause everybody to come safely to a stop and potentially do an all-road extension if they go through the red light. But so far we're not doing that. The higher traffic intersections where we're deployed right now don't seem like a great fit for that and speeding is not really an issue there.

Ali | Startup Explore (12:19)
I see got it. And given that your system has a direct impact to public safety, how do you guys ensure that the system software is always accurate and always correct and reliable?

Collin Barnwell (12:33)
Yeah, great question. we're communicating to the ⁓ traffic controller, is sort of below that. There's this thing called the, what's it called, the MN?

⁓ which is sort of making sure that nothing terrible happens, there's no like green lights in every direction type of thing or...

Ali | Startup Explore (12:57)
Mm-hmm.

Collin Barnwell (12:58)
If nothing is coming in and everything's offline, ⁓ you know, it's just red flash, right? So that's, that's sort of the base layer of safety outside of our system. Then there's the traffic controller, which has a timing plan programmed into it. So that's minimum green light, maximum green light. ⁓ How long is the yellow light? How long is the red light? And we're, following that timing plan. So we're not overriding that in any way. The only kind of way that we would.

Ali | Startup Explore (13:16)
Mm-hmm.

Collin Barnwell (13:27)
deviate from any of those values would be our all red extensions. ⁓ But we're doing that in a pretty well defined way. And so if we kind of go offline, something bad happens. The controller is still operating that way.

Now that's still suboptimal because we need to be sending in detection inputs. And so one layer above that, we have a separate much smaller piece of hardware that mediates our connection to that traffic controller. Where if our main GPU system just goes offline or is going through an update or something is happening, that system has a fallback mode where...

it is still sending detections under certain circumstances to the controller. And then all the way up at the top of the stack, like where our actual software is running on the GPU, we just test that a lot before we ship anything. We're constantly monitoring everything. We can see everything that's going on from our office to make sure things are going well. ⁓ But yeah, we have a lot of layers of fallback to make sure things are working.

Ali | Startup Explore (14:37)
Makes sense. then the traffic controller, is that kind of a hardware piece that is operated by the city itself, the government that has nothing to do with you guys, right? Okay. Okay. And when it comes to ⁓ thinking about the installation and integration to existing traffic lights, how easy or difficult it is and what's the usual timeline.

Collin Barnwell (14:47)
Correct ya.

So there's two main standards. In the US at least, there's a Caltrans standard, which was created by the California Department of Transportation, but it's used in a couple other states. And then there's a NEMA TS2 standard. The NEMA TS2 standard has one or two different protocols for communicating with the controller. And the Caltrans, all the controllers are basically the same, and so there's one way that we do

We just need to implement those three protocols basically and we can kind of work with any controller. If there's a controller that we haven't seen before, we'll make sure that we don't need to do kind of a new integration. But for now, it's just those two protocols basically.

Ali | Startup Explore (15:52)
And sorry if I missed, but in terms of the timeline, how long would typically take to integrate?

Collin Barnwell (15:57)
I mean,

we were typically not the slow ones, right? We're kind of ready to go next week. If the city is, ⁓ we will sometimes, ⁓ you know, talk over the timing plan, make sure that, you know, all the values in that time plan are going to work with the system. ⁓ but it's really about the actual physical deployment. That's kind of the thing that.

I guess the sales cycle and then the physical deployment. So we don't actually go out there and if necessary, hang up the cameras. The city will contract that out. So we're waiting for that to happen. And then the city electricians will install our hardware unit, ⁓ plug it into the controller, configure the controller to accept those inputs. And that, yeah, that kind of happens on their timeline.

Ali | Startup Explore (16:32)
Mm-hmm.

I see.

Got it. With that being said, I guess, how many pilot programs or deployments you have currently in the cities? And of course I'm mindful that you guys are still pretty early stage company and early in the...

Collin Barnwell (17:07)
Yeah, so our

first pilot, our first, not pilot actually, our first deployment went live in San Anselmo, California ⁓ in the beginning of August. And ⁓ we are...

where our next our second intersection is going live tomorrow ⁓ also in San Anselmo.

Ali | Startup Explore (17:28)
Nice.

Collin Barnwell (17:29)
And the city has applied for a grant to go citywide with our system. ⁓ That should be happening over the next few weeks or months if everything works out well.

Ali | Startup Explore (17:43)
Awesome, congrats. And what are some of the metrics that you're getting from that first deployment? And to add to that, what key metrics you guys try to measure and track?

Collin Barnwell (17:57)
Yeah, so we're tracking ⁓ wait times is sort of the key number ⁓ from a safety side because these kind of negative events are so rare. ⁓

We're tracking like safety features at this point, rather than actual outcomes. Although in the longterm, we expect to see the impact on outcomes, just pretty low end right now. so we're tracking wait times at this intersection, that first one where we went live just to kind of set the scene. This is the most congested intersection in Marin County, according to a recent traffic study. Marin County is a county just across the Golden Gate Bridge from San Francisco. so it's called the hub in San Anselmo.

and we're seeing 30 % reductions in rush hour wait times. And then we're also on top of that supporting these additional safety features like dilemma zone protection and dynamic all red extensions, which, you know, if we got rid of those, we would improve wait times even more, but we want to kind of be improving both of those things together.

Ali | Startup Explore (19:03)
given that your AI technology processes a large number of vehicle and ⁓ pedestrian data, how do you approach privacy of that data?

Collin Barnwell (19:14)
Yeah. our agreement with San Anselmo says that we are not collecting faces or license plates. ⁓ so we, we have an LTE connection to the, ⁓ to the device in San Anselmo that is not.

anywhere close to enough bandwidth to actually be streaming video data at any ⁓ kind significant bandwidth to the cloud. We do use some of our video that we go out there and manually download for improving upon the algorithms and checking that things are going well and ⁓ improving the object recognition. But we can't see license plates or faces.

Ali | Startup Explore (20:01)
I see. Now, from your journey so far from building and scaling in the company and then landing first customer, what's been the biggest challenge?

Collin Barnwell (20:12)
⁓ definitely in a market like this, is, ⁓ cities understandably like want to get it right. ⁓ it was pretty hard to find someone who wanted to be the first place where this was deployed. ⁓ so fortunately we, found that person, in San Anselmo.

I think we found a person who had a huge need, which was like, this is the most congested intersection in the County. There were some solutions identified to improve that intersection that were really, really, really expensive. Uh, like, you know, $50 million to build an overpass, $25 million to build a roundabout those, those sorts of solutions. And we came along offering something at a small fraction of the price.

So I think a combination between, um,

you know, a clear need and also, ⁓ you know, props to our first customer, Scott Schneider in San Anselmo, a willingness to work with some pretty early stage, you know, founders who are very interested in solving this problem. But, you know, it's, it's, there's a lot of back and forth and getting something like this off the ground. So, ⁓ yeah, the hardest part was, finding, finding that first customer.

Ali | Startup Explore (21:40)
Yeah, I hear you. And I think for a lot of the people out there who are about to start the company or early in the stages, landing that first customer is one of kind of a big milestones. And let me double click on that and ask you, so how did you guys find that first customer? What actions you took? As you can provide more color there.

Collin Barnwell (22:05)
Um, yeah, I called, uh, someone in Nevada, which is a different city in Marin County, um, just called to see if they had any kind of traffic issues and if we could be helpful. And, uh, I got someone on the phone who was not that interested in talking to me. kind of seemed like, and, uh, suggested that instead I go and present at the Marin.

Works Association meeting, ⁓ which I think that in that situation he was sort of like he didn't really want to, he wasn't that interested but he was sort of like maybe you should do this and get off the phone with me. ⁓ But I took that as like that's a great idea and so I reached out to the organizer of that meeting. We went, my co-founder and I...

to that meeting in San Rafael. We presented to a room full of public works employees and one of those was Scott Schneider in San Enselmo and he was super interested in the technology that we were showcasing and you know we continue the conversation from there.

Ali | Startup Explore (23:18)
Nice. Did you have those like data points at the time that when you were kind of trying to lend that customer around, Hey, overpass would cost you this much. Roundabout would cost you this much. Our poll, our solution is way cheaper or those data points came to you later.

Collin Barnwell (23:36)
Those data points came to us later.

Ali | Startup Explore (23:38)
Got

it, got it. Okay. Cool. Now, let me take you back. Can you talk about your and your co-founder's background,

Collin Barnwell (23:46)
Yeah. ⁓ so my background is I worked on data center networking at Google, kind of low level software before I moved on to Verily Life Sciences, where I was working on medical devices. I've built various kind of physical, physical devices that have software on them. I've built implanted nerve simulators and retina cameras and, smartwatches and that sort of thing.

And my co-founder has spent his whole career in the self-driving car industry. So he's worked at Zucs and Waymo and applied intuition. ⁓ We've been friends for like eight years. Mostly, I guess, got to, we have a really good friend, a mutual friend that introduced us when I first moved out to the Bay. They were roommates in college and then I lived with this guy when I moved out here. ⁓

And ⁓ yeah, we both kind of had an interest in housing and transportation politics and that sort of thing. So we kind of hit it off in that way. And, you know, we've been friends for a while and I was poking around the space and eventually convinced Sabik to join me on this venture.

Ali | Startup Explore (25:05)
Nice.

And I believe you both are pretty technical and did you both majored in computer science?

Collin Barnwell (25:12)
Yeah, we both did. Actually, no. He studied economics undergrad, but he got a master's in CS. Yeah. Not the only econ major CTO that I know too.

Ali | Startup Explore (25:18)
Okay.

interesting. Got it. Okay. And ⁓

That's a pretty unique skill set in Blend, guess. Interesting.

Collin Barnwell (25:33)
Yeah, I don't know.

It's, yeah, it's a thing, so. ⁓

Ali | Startup Explore (25:37)
Nice. ⁓ And you guys went through YC program, is that right? ⁓ What was the, maybe the biggest highlights from that experience?

Collin Barnwell (25:41)
Correct. Yeah.

It was a really good experience. like, I have a lot of respect for YC as an institution after going through that program. ⁓ So there were a lot of highlights and a of things, but I think that the enduring thing has been just the community that has a really high bar and a really high...

like belief for what's possible, I guess. So we meet with a couple other people from our cohort every two weeks to set ambitious goals and see how we do on them two weeks later. And when you do that day in and day out for, it's been over a year now, meeting with that same group of people. ⁓

People start to take off, you know, and it really shows you what's possible and ⁓ it really shows you what works and what doesn't work. And you also got to see them when they weren't really taking off. And so it also kind of shows you that it doesn't start off feeling that way. So I think all those things just sort of normalize the...

I guess all the crazy stuff that you have to do to get a startup off the ground.

Ali | Startup Explore (27:05)
It's actually nice. I never thought that you guys, after graduating from YC, you would still meet on a regular basis with the cohort that you've graduated with. That's nice. That's awesome.

Collin Barnwell (27:16)
Yeah. I think it's,

obviously it's kind of a self organized thing. Like it's an optional thing. ⁓ but I know that they, kind of encourage it and I think probably most people do. So.

Ali | Startup Explore (27:23)
Mm-hmm.

Nice. Now, switching gears a little bit and thinking about total addressable market for you guys. I'm pretty sure it's huge, but help me to understand how to even define the total addressable market for you guys. Is it the number of intersections globally or is it something else?

Collin Barnwell (27:52)
Yeah, so we kind of think of it as there's 330,000 intersections in the US. ⁓ We're early. We're still kind of figuring out our long-term pricing scale or long-term pricing strategy ⁓ and obviously international. ⁓

Traffic signals is a whole nother market that we can go after. ⁓ We just kind of look at it as 330,000 times whatever we can charge is big enough to warrant us being a company in this space and we'll figure it out from there. I think more fine-grained TAM calculations than that are VC's job not.

Ali | Startup Explore (28:23)
Yes.

Collin Barnwell (28:32)
Not really ours. And then we'll just, we'll just put one foot in front of the other until, until it's a giant company. But I think it's like, the answer is it's big enough.

Ali | Startup Explore (28:34)
Yeah, makes sense. hear you. mean...

Yes, yes. I'm pretty sure it's big enough and I think it's going to be getting bigger and bigger as we go. ⁓ Cool. Now talking about competition, do you have that or in that competitors in the space? Competition in the space.

Collin Barnwell (28:59)
Do we have what?

yeah. so there are a lot of companies that do, like I kind of explained in the introduction, like the detection of the presence or absence of vehicles. So on some level we, we compete with companies like Econolite and Cubic and Myovision that like give you the ability to detect the presence or absence of vehicles. because we're replacing those systems,

Ali | Startup Explore (29:11)
Mm-hmm.

Collin Barnwell (29:31)
there are competition. But our capabilities are pretty different. A little bit more similar to these systems called adaptive systems, of which there's a couple that use, historically they use kind of the same input data, so presence or absence of vehicles in specific locations, to make more complex decisions. So they'll slowly...

Increase the length of green when there's more traffic decrease the length of the green lights when there's less traffic ⁓ But they're they're doing that based on this count data Which is sort of backward looking if you think about it. You're detecting the presence or absence of vehicles as they pass by the intersection You're looking back in time. You're not looking forward at what's coming. And so ⁓

And so those systems tend to have issues where they either over-correct or they under-correct, and those are parameters that you have to tune and get right so the implementation of these systems can be somewhat complicated. But that's kind of what the space looks like.

Ali | Startup Explore (30:41)
Got it. So I guess to let me make sure I'm on the same page and correct me if I'm wrong, but what sets you apart is basically your ability to project the car speed, whatever the path versus the competition. They only look backwards into the data and they don't do as much of that projection. And that makes me think you probably use more data points.

than maybe some of those competitors in your ⁓ software. Okay.

Collin Barnwell (31:12)
Yeah, exactly. So

we're mapping everything in 3D space. We have position and velocity, and we're using that to kind of run simulations 10 times a second to figure out what the best signal change is, both for traffic and for safety.

⁓ Other adaptive systems would be looking at...

the actuation of these simple binary detectors to try and make the same decision, which it's hard to nail that. And so they have a bunch of other parameters on top of that. ⁓ If somebody needs to figure out what those parameters should be, they can get them wrong. ⁓ The detectors can be wrong about something and throw off the algorithm wildly. yeah, does that make sense?

Ali | Startup Explore (31:39)
Mm-hmm.

Mm-hmm.

It does. does. thinking and looking into the future, let's say five years from now, what is the North Star for the company? Where or how you think this technology will evolve? And where do you see your roundabout in that space, especially given ⁓ the this autonomous cars ⁓ picking up and then maybe cities becoming more smarter and smarter? How do you see your roundabout to being kind of a

Collin Barnwell (32:24)
Mm-hmm.

Ali | Startup Explore (32:32)
part of the puzzle in five years.

Collin Barnwell (32:35)
Yeah. So, I mean, five years from now, I would hope that we are support other things that I've mentioned here. and hopefully we found a lot of other ways to be improving road safety. So I think I've, ⁓ talked about like dynamic red extensions in various situations. ⁓ but if we can respond to more interesting or complicated situations, like blatant red light running,

we can kind of proactively detect and predict that and respond to that, that would be awesome. I think from the autonomous vehicle perspective, like there is this whole protocol called VDACS that... ⁓

It's a communication protocol between basically like traffic lights and vehicles. ⁓ We may or may not choose to implement that protocol at some point in the future. Our thinking right now is sort of.

We can see everything that this protocol communicates. You can have a protocol that says I'm approaching this light from over here, but we can also see that pretty well and pretty accurately. ⁓ We could, if it's useful to self-driving car companies, implement protocols to broadcast certain forms of data, ⁓ but I think that's kind of down the road for us.

Ali | Startup Explore (34:02)
And as you grow the company, you'd need to hire more and more people. I'm curious to ask, how do you approach hiring and what are specific qualities that you look for in candidates?

Collin Barnwell (34:15)
That's yeah, that's a good question. We're quite small right now. We have four people, including the two founders. one of those people is very close friend of both of us that is super smart. And we just like, you just know, you know, when you see it, and then the fourth person, ⁓ is someone we posted a

job on the Y Combinator job board, he applied. I interviewed ten-ish people, just like in English, like conversation, like trying to subject matter expertise and you know, just like are they, are they gonna be a good fit for the team and...

Ali | Startup Explore (34:54)
Mm-hmm.

Collin Barnwell (35:03)
You know, with, like with AI, kind of sort of feel like it breaks the traditional interview process a little bit. ⁓ so that was, that was important that it's just like, I need to, I need to know that this person knows what they're talking about. and then we did do a coding interview to make sure that they were competent, writing code, ⁓ and they were, and that person has been also an amazing hire. So.

Ali | Startup Explore (35:26)
Mm-hmm.

Collin Barnwell (35:32)
On some level, like we've hired once and it went really well. And also I should say that we like open the job. We got a ton of applications. We interviewed immediately. We hired immediately and the whole process was done in like a week and a half. So I don't, I, I'm told that hiring is usually harder than that. Um, but that was not our experience. So.

Ali | Startup Explore (35:43)
Bye.

Yeah, I feel like when you build something impactful as roundabout, I mean, has direct impact to your everyday lives of so many people. Like people get a lot excited and would be very much interested to join. Awesome.

Collin Barnwell (36:09)
Yeah. I thought,

yeah, that has kind of been our experience. guess it is cool to work on something that's physical in the real world. something that everybody interacts with on a daily basis. And honestly, that was kind of selection criteria for figuring out what I wanted to work on. I don't think, I don't think I could have convinced Sabik to work on something any less interesting. yeah.

Ali | Startup Explore (36:26)
Mm-hmm.

That's awesome. Yeah, it's pretty exciting stuff. Now question outside of work. Can you tell me what you like to do? What kind of hobbies you have?

Collin Barnwell (36:45)
Good question. have an eighth month old baby and a start. Thank you. ⁓ But those two things both take up a lot of time and I have not had a ton of time to do like my typical hobbies on top of that, especially since my son was born. So ⁓ I do love skiing. like pretty much anything that will get me outside. ⁓

Ali | Startup Explore (36:52)
Congratulations.

Collin Barnwell (37:14)
I surf a little bit. ⁓ I love biking. I guess I still have time for biking because that's how I get to work. So that's nice. ⁓ And I like playing music, play guitar and bass, ⁓ sing a little bit. So yeah.

Ali | Startup Explore (37:18)
place.

That's awesome. Do you want to give it a try here for singing? That would be interesting to listen. And by the way, what's your kid's name?

Collin Barnwell (37:36)
Nope. Aghi, Agustin.

Ali | Startup Explore (37:46)
Augustine, very nice, very nice. Awesome, congrats again. Awesome, Colin, I know we've kind of talked a lot about the company, about the team, the future vision. there anything else that you'd like to share that we haven't touched on?

Collin Barnwell (37:49)
Thanks, Ian.

⁓ If you or someone you know works at a city and is interested in having dramatically more efficient lights with awesome safety features, please email me and I will get back to you very quickly.

Ali | Startup Explore (38:19)
Yes.

Yeah, again, I mean, it's a pretty exciting project. And I feel like a lot of people who might even watch this video might be interested to kind of give you those leads and ⁓ we can put like the contact email at the description so people can check out and ⁓ come up with a few leads at least that'd be great. Well, ⁓ Colin, thank you so much for coming to the show. It was really exciting to learn about you about the company.

and I hope that we will meet sometime soon again and you will have even bigger successes to share.

Collin Barnwell (38:52)
Absolutely, yeah, thanks a lot. This was a lot of fun.