CARMERA: Providing Smart Data to Self-Driving Cars
Dear fantastic readers,
For our second week, we are blessed to present the insights of Ro Gupta (’00 ORFE), who is the founder CARMERA, an autonomous vehicle technology company that runs a crowdsourced vehicular sensor network to gather continually updated, three-dimensional spatial data and street analytics. Prior to that, he was a VP of Business Development @ Disqus (10mm -> 1.5 billion MAU), as well as an advisor to number of different startups.
How did you get the idea for CARMERA?
I think it was a confluence of many different things. I remember having a conversation with Kyle Vogt, founder of Cruise Automation, back in 2014. He was working on retrofitting cars with technology to make them autonomous. One of the projects he was working on was using the data those cars collected for mapping, and I thought, “I wonder if that data can be collected another way? Could we crowdsource this?”
About a month after that conversation, my wife and I were stuck in Labour Day traffic on the way home from a trip, and I thought, “Wouldn’t it be cool if all these cars were equipped with dash cams, and you could see through other people’s dash cams?” If you’ve ever seen the “car periscope” episode of Curb Your Enthusiasm, that’s kind of what we were wishing for! The original idea was a sort of internet-connected dash cam product. That evolved into the model we have today, where we gather data with fleets of our own autonomous vehicles and with our Fleet Monitoring partners.
Your most important product is the CARMERA autonomous map, which provides real-time data about the streets to your clients. I can only imagine the enormous quantities of data you must be collecting for this task. How do you go about cleaning this data, sculpting it, and displaying it in a useful way?
Our process is very layered and it has a lot of moving parts. We start with a high-quality base map (made out of 3D data gathered using LIDAR), on top of which we layer on semantics, which includes features like lanes, street signs, etc. But all of that’s still static. Then we add change detection, the dynamic layer. All our data pipelines are modular and reconfigurable to serve different clients.
Was there a specific moment in the development of CARMERA when you thought, “Wow, my idea is really going to work?” What was your first real taste of success?
There’s this mythology of “the overnight startup” that doesn’t really exist, but there are a few “a-ha” moments I can name.
One huge confidence boost right at the beginning was doing the homework, validating and invalidating ideas before even building anything. The dash cam idea was invalidated pretty quickly. We talked to a lot of drivers, and it seemed like the interest in such a product just wasn’t there. But then we started talking to professional drivers, then their bosses. I realized that it was their needs, not the needs of the general consumer, that lined up well with what we could deliver.
The more exciting moment was when we got a team together, raised our first round of funding, built the sensors, and convinced our first fleet to install the sensors. We weren’t sure how useful the data would be. It wasn’t until weeks later when our engineer Chase built a UI that showed the video data, and I remember thinking, “That’s gotta be wrong, I don’t see how we could’ve covered that much ground.” But it was accurate! It showed how much ground we could cover with just a handful of vehicles, and that we were getting useful data.
But after that, then what? We still had more to do. Another round of testing, funding, uncertainty. There’s always more on the path ahead of you.
You mentioned in a Keller Center interview that Alain Kornhauser and Ed Zschau are among the faculty here at Princeton that planted the seeds of entrepreneurship in your brain. Can you tell us more about them and how they affected you?
Kornhauser inspired me on the subject matter. Back in 1998, we were essentially modelling autonomous Uber (granted, we thought it would be with monorails, so we missed some minor details). Ed Zschau’s class, High Tech Entrepreneurship, was what got me interested in startups. I was always interested in being an inventor (by the way, ideas mean nothing), but Ed exposed me to the startup part, and broke down the concept into manageable parts. Ed was also just a very inspiring teacher. Lots of people point to his class as the reason they’re interested in entrepreneurship.
Is there anything else you would like to tell the Princeton entrepreneurial community?
Ideas mean very little, execution means everything. People think they’re owed something if they think they had an idea first. But if you actually look at some of history’s most important inventions, it’s almost unheard of that only one person thought of it. One person gets the credit, but if you dig deeper you’ll find that lots of people were working on it at about the same time. The difference between them was execution (and sometimes luck). So much of the value ends up coming from the execution, the testing, the building, the marketing, etc. That’s what actually breeds success. Ideas are good, but ideas alone are nowhere near good enough.
In the previous interview I did with the Keller Center, I said that if you’re thinking of doing something entrepreneurial, it’s okay to wait a bit if now doesn’t feel like the right time. But if you’re weighing the options and it’s an even tie, go for the riskier option. One of the thing that Jeff Bezos talks about is the regret minimization framework: making decisions based on what you will least regret later in life. If you talk to anything in their twilight years, most of them never say, “I wish I played it safer.” Most of their regrets are “Oh, I chickened out.” I think there’s so much wisdom in that. My idea was nagging at me enough that if I don’t at least try it, it would’ve driven me crazy! I wanted to see it go somewhere, or see it fail. If you feel that, you have to give it a shot. These days that’s easier than ever, because the perception of startups has changed. Risk-taking, even risking and failing, is no longer criticised, but admired. Having Princeton on your resume means you’ve already got “social proof” of success. That is a big advantage. Take it as an incentive to go on the less-trodden path.
CARMERA is currently recruiting several full-time positions, and it is also participating in this year’s Princeton Startup Immersion Program. Find out more here and here.
— Shout-out to Matt Yuan ’21 on this great interview :)
That’s it for this week. As always, if you have any feedback or are involved with entrepreneurship in any way, role, shape, or form (VC, engineer, founder, etc.) we’d love to talk to you :)