Recogni is enabling ADAS Level 2+ and Autonomy with its Peta-Op class, high-performance and low power consumption vision inference system. The Cupertino, California-based company, was founded in 2017 and counts Toyota Ventures, BMW i Ventures, Faurecia, Bosch and Continental among its investors.

We are a “Silicon + Systems + AI” company”, R K Anand, Founder & Chief Product Officer, Recogni

We spoke to RK Anand, Founder & Chief Product Officer at Recogni on:

• Why computing is a key enabler for safer ADAS, as well as higher autonomy;
• Recogni’s advantages in high-performance, low-power consumption computational platforms for perception processing;
• How product fit and customer acquisition matter for Recogni’s growth strategy;
• Recogni’s success story, milestones and growth velocity

To listen to the full interview join Auto2xperts.

Contents

1. About Recogni

• Capabilities as a “Silicon+Systems” company, across the U.S & Germany
• Product, market fit & customer acquisition: top priorities today

2. Recogni’s product differentiation

• Recogni’s advantage: high-power, low-consumption computing platforms
• How to unlock full autonomy: Sense (Perception) | Plan (Computing) | Act
• Computing has been a key challenge in upgrading ADAS camera resolution and frame rate for safety and higher autonomy
• The industry needs to shift from low-performance sensors that merely comply with regulations to perception accuracy that saves lives

3. Product positioning and market fit

• Optimised design for low power delivers cost efficiencies and scalability
• OEMs who are falling behind in Autonomy could benefit from Recogni’s chips

4. Growth trajectory

• Leveraging the technological advantage to increase customer acquisition
• Milestones: Deployment in 2023 with Trucking customers and 2024 with Passenger car OEMs
• Product roadmap: integration, next-gen chips & systems

5. How to overcome the challenging road ahead

• Alignment with client delivery schedules is crucial
• Scaling up the team by complementing compute experts with experienced automotive teams

The need for computing to advance driver assistance systems to highly autonomous driving

Autonomous Driving computational platforms that combine high-performance processing of vehicle perception data and low power consumption are one of the building blocks to transition to higher levels of autonomy. Today, many players are facing challenges with real-time processing of the vast amount of data required to enable SAE Level 3 and 4. Higher levels of autonomy require enhanced perception redundancy, new centralized architecture, and driver-facing cameras among others. The new software and hardware requirements will drive demand for new generation ADAS sensors, chips and super-computers, AI, HD maps etc. They will also drive further collaboration between OEMs and Tier 1s-2s for the development of AD platforms.

“The technological advancements of Recogni’s computing platform could solve the perception processing challenges that hold back the transition from ADAS to Level 3-4 autonomous driving”, says Auto2x

Capabilities as a “Silicon+Systems” company, across the U.S & Germany

“Recogni has a presence in Silicon Valley, U.S and Munich, Germany therefore, it’s a bi-continental company. We are venture-funded and strategic-automotive-funded. On the venture side, our investors are Great Point Ventures, which led the Series A round, and then Celesta Capital which led the Series B round. In addition, we have Mayfield as a venture investor. On the automotive strategic side, our investors are Toyota Ventures, BMW i Ventures, Faurecia, Fluxunit, Bosch and Continental. We’ve raised $74.8 million over Series A and Series B (Feb’21). We are a “Silicon + Systems + AI” company, i.e. our silicon and systems teams are here in California, whereas our AI team is in Munich, Germany. So, that allows us to be bi-continental in nature. Also, it allows us to execute across multiple dimensions of engineering and product development.”

Recogni’s product differentiation

“Our vision for Recogni was to solve the fundamental compute problems in vehicles for AI. Our purpose was to build compute that is an order of magnitude better than anybody in the world. Two of the most important metrics in computing across the industry are the maximum performance usable TOPS and the absolute compute. We are at least 40 to 50X TOPS, a lot better than anybody else in the industry. We also have the highest level of computing. We have a functioning chip, and a functioning system, and object detection or semantics segmentation features are being tested. Technological innovation and the ability to hire the right kind of talent give us those advantages. We have the world’s highest-performing AI inference device at the edge for vision processing, the first Peta-Op class (one quadrillion operations per second) device. Our chip is at least 8X, maybe 10X, better than anything else in the industry. It also has a power consumption number that’s lower than the industry’s leading devices. We’ve not only built the chip and the software but have a functioning system that operates at a better op/s at very low power at 25 Watts. The combination of computing performance and low power consumption allows us to achieve sensing capabilities that are getting close to real human perception.

We intend to build upon our technological advantages and maintain our performance leadership with next-generation designs and systems”