Introducing DeepMap:

The Team Helping Make Autonomous Driving Safe

By James Wu and Mark Wheeler, DeepMap Co-Founders

We are excited to introduce DeepMap, our new company focused on scalable and accurate mapping and localization for autonomous vehicles.

Over a year ago, we started stealthily building a team of some of the best engineers in our field, developing our technology, and raising two oversubscribed rounds of funding. With the closing of our Series A last week, it was time to share more about what we are doing.

Some history

While the autonomous driving space is fairly new, the two of us have been working on the underlying mapping and cloud infrastructure technologies for decades.

One of us (James) led the engineering efforts building the serving infrastructure of Google Earth, helped with the launch of Apple Maps, and served as principal architect for Baidu’s self-driving platform.

The other (Mark) started early, conducting “localization” research at Carnegie Mellon University, worked at Apple on panoramic imagery, made breakthroughs in LiDAR technology and point cloud processing at Leica Geosystems, and built enterprise cloud mapping solutions that served petabytes of data at Google.

And we are fortunate to be part of a team of similarly experienced software engineers. Collectively we have built, scaled, and maintained mapping technologies used by billions of people, including some of the world’s largest and most successful mapping platforms like Google Earth, Google Maps, Google Maps Engine, and Apple Maps.

Mapping for safe autonomy

Reflect for a moment on driving a very familiar route — from your home to your office, for example. You already have a mental “map” of your commute before you begin driving, making it easier to focus on the truly safety-critical parts of the drive. For instance, you can anticipate unusual driver avoidance behavior where a large pothole has been for weeks, and know the speed limit despite signs being blocked by a large truck.

Now compare this to driving a completely new route, when you have much more information to process because everything is unfamiliar. You can only react to what you see in the moment. For example, you may miss a highway exit because the exit sign was blocked by a truck next to you, you may fail to slow down for a school zone, and you have no way of correctly anticipating unexpected road conditions. Or you may turn the wrong way into a one-way street, not realizing the change in traffic patterns until it’s too late.

The same principles apply to self-driving vehicles. High-definition 3D maps make routes familiar to self-driving cars, which in turn makes them safer.

Traditionally, maps are seen as “databases” that are sold to customers, with periodic updates. Then, it’s up to those customers to figure out how to make the best use of the map data.

Whereas today’s maps are made for humans, maps for self-driving cars are made for machines. This different end user presents new challenges, ranging from difficulties keeping maps fresh to lack of performance guarantees during the map consumption phase, such as real-time localization capabilities.

We think about mapping differently: as a full-stack service that optimizes the entire process from map creation, to map consumption and map serving. To ensure map quality and consumption performance, maps need to be made from self-driving cars and for self-driving cars, creating a virtuous cycle.

DeepMap provides a high-definition mapping and localization service designed to support millions of cars while keeping map quality high, map consumption highly efficient, and cost very low.

Self-driving cars using our service will know precisely where they are on the road, what’s coming around the corner, when and where they are allowed to make turns, and how conditions might have changed since the last time they drove this stretch of road — and all with extreme efficiency and low cost.

Our full-stack mapping and localization service scales smoothly in line with the growth of self-driving fleets. It turns our customers’ self-driving fleet data into high performance maps — your data, your maps.

Without this kind of full-stack mapping and localization system, self-driving vehicles will not be safe enough for true autonomy. This is a critical piece of the puzzle needed for the ambitions of the self-driving industry to be fully realized.

Raising $25M Series A round

All of this brings us to today’s news: we’ve raised a $25M Series A round of funding led by venture capital firm Accel along with existing investors Andreessen Horowitz and GSR Ventures. This funding adds to our previous seed round of nearly $7M that was led by Andreessen Horowitz.

We’re excited for this new chapter in our business, and are looking forward to continuing our mission of solving the mapping challenge for autonomous vehicles.

We are confident about our vision and believe that our team is well positioned to help accelerate the rollout of safe self-driving technology. We will have a lot more to share in the coming weeks and months, so stay tuned!