About Us

Recogni’s mission is to design a vision-oriented artificial intelligence system from the ground up. The system will deliver unprecedented inference performance through novel edge processing, allowing vehicles to see farther and make driving decisions faster than humans while consuming minimal amounts of energy. The company is headquartered in San Jose, Calif. with operations in Munich, Germany and is backed by leading venture capital firms including GreatPoint Ventures with participation from Toyota AI Ventures, BMW i Ventures, Faurecia, Fluxunit – OSRAM Ventures, and DNS Capital.

Our History

Recogni was incorporated in 2017 to help auto manufacturers reach Level 2+ vehicle autonomy, but our roots go way back to 2013. Ashwini Choudhary, one of our co-founders, grew up riding motorcycles, starting when he was around 12 or 13 years old. Bikes have been an integral part of his life, and even though he’s not allowed to ride anymore, he still loves them.

It began with a beautiful Sunday ride with friends in the backroads of the Bay Area. In the early evening, he was on the final stretch of “twisties” to get home, maybe 3 miles away. There was a car coming in the opposite direction on this country road. The driver must have been texting or fiddling with something, and she hit him while making a left turn. 

Three weeks later he woke up in the ICU, needing over 12 surgeries to put him back together. 

Ashwini spent the next seven months in various hospitals and then some quality time in rehab. Ever the eternal optimist, it was during this time that he kept returning to a single idea: how do we prevent these accidents from ever happening again? There had to be a way for the vehicle take control, especially when the driver was distracted.

This idea would become Recogni.

After rehab, Ashwini sat down with both RK Anand, an old classmate, and Eugene Feinberg, a fellow co-founder of the camera startup that he and Ashwini were working on together. Both of them were polite and patient, and these faults of theirs gave Ashwini enough time to convince them that this was a worthwhile journey. Together they discovered the real market opportunity, and Ashwini was able to get them on board to build a company around revolutionizing safety and autonomy. 

Examining existing safety systems, they quickly realized that the current batch of accelerators could not be scaled to solve the mobility problem beyond Level 2 autonomy. This is because Level 2+ (L3/L4/L5) autonomy requires exceptionally high realtime processing performance while consuming little to no power. 

Enter Recogni. Our goal is to change the trajectory of the autonomous vehicle market, paving the way toward fully autonomous and safe driverless vehicles. It’s a tough road, but nobody ever said it was going to be easy.

What makes Recogni different

Energy Consumption

Able to compute at 1000 TOPS while consuming just 5 watts (chip level)

Visual Range

Recognizes small objects (20 cm x 20 cm) to 300 meters

Performance

Per chip computation can scale to 1 peta operations per second

Power

12W including sensors passively-cooled (system level)

Efficiency

Over 75% compute efficiency in typical ML based vision application

Architecture

Module designed around proprietary chip for inference acceleration (perception processing)

Sensor Frontend

3 integrated image sensors (RGB, IR, Monochromatic) and interface to external depth sensor (LiDAR or RADAR)

Output Intelligence

Outputs tensors to central system with object recognition (detection, segmentation, classification) and other properties integrated from depth sensor