Reshaping the Machine Learning Landscape at the Embedded Edge

by Krishna Rangasayee

At, we believe that the future of compute is high performance machine learning (ML) at the edge – and today, power is the limiter. We are passionate in enabling our customers to build green high performance machine learning solutions at the embedded edge across diverse industries.

Today, we are closer to achieving our dream as I am very excited to share that, is raising an additional $30 million in our Series A round. It is a key milestone for our employees, customers, partners, and investors. After just 15 months of starting the company, we are working with over 20+ industry leading customers in the areas of robotics, smart cities, autonomous vehicles, government and medical sectors. I feel immense gratitude towards our talented team for executing on our vision focusing on our customers needs and bringing their passion to work every day.

Listening to customers from the beginning

Right from day one, we have been engaged with our customers asking, “What are your biggest challenges and how can we help?” Before we did any innovation or execution, we spent time listening to develop a deep understanding of our customers’ system-level problems and needs. We learned that companies want to derive ML benefits without a steep learning curve, while simultaneously preserving their legacy IP and working in a very easy to use software environment. At the same time, companies are struggling to address power issues, especially when high performance compute is required. These issues were critical gaps across almost every major customer we spoke to. This learning shaped our outlook and provided clarity on the computer vision system-based problems that needed to be solved at the edge. We also set out to build an architecture and early version of the software to prove to those customers who invested so much time with us, that we could solve their compute performance needs while simultaneously making it dramatically low-power and easy to use.

The collective learning from our customers helped us focus on solving complicated system level problems with a heterogenous compute platform that simultaneously supports traditional compute with a ML environment, providing a seamless software experience for our customers. We picked the space of computer vision as our first area of focus and set our targets to solve some of the industry’s biggest challenges. Our systematic and disciplined approach brings us to this step in our journey where we are now deeply engaged with over 20 customers, who are leveraging our automated tool flow to run their own applications, our MLSoC™ architecture is now frozen, and the nonlinearity of the architecture development and proving our software approach is behind us.

Dream team takes risks to solve tough technology problems

Based on all the customer feedback, we built an architecture that was software centric, easy to use, and consistently 30x better in frames per second/watt (FPS/W) than alternatives. We spent a lot of time handpicking the talent to help solve the problem. Our needs were unique in wanting to build a software-centric company that supports both traditional compute and ML environments in one integrated product offering. We assembled a founding team of talented people who brought great strengths in ML, ML software, SoC development – all of them with a strong background in innovation and execution. The combination of all of their talents gave us the best shot at building a very software-centric approach to address the problem of delivering high performance ML platforms ranging from 25-400 TOPs at a 10 TOPs/W efficiency. We then expanded the team and have now grown into 35+ amazing people who were previously with industry leading companies. The team has cumulatively delivered 50+ tape outs, 10+ compilers, and developed 10+ architectures over their career spanning a wide range of technologies across SoCs, CPUs, GPUs, and FPGAs. We broke many rules and put the end goal above any one person’s opinion or approach and enjoyed the process of building a brand new product category that translated into what is now MLSoC™

Fulfilling the technology promise, delighting our customers’s Machine Learning SoC (MLSoC™) is the first machine learning platform to break the 1000 FPS/W barrier with >30x improvement over alternative solutions. The MLSoC platform is reshaping what was previously held possible and supports traditional compute with high performance machine learning, lowest power, safe and secure machine learning inference. In close partnership with our customers and key technology partners, we now have the architecture in place and with our software tool chain, enabling customers to engage with us and delighting them on what they can do. We are so proud of our technology advancements to date and excited about reshaping what is possible in machine learning. Our expanded team, who I deem as our extended family, is innovating and executing to the architecture and driving us to production.

Our Series A financing gets us closer to fulfilling the technology promise we identified along with our customers from the onset. Initially, focused on helping our customers drive amazing solutions for computer vision applications at the embedded edge, our team of technology experts and engineers are committed to delivering the industry’s highest frames per second per watt solution and starting a technology movement centered on the theme: Is your ML Green?™

Backed by Top Investors

Dell Technologies Capital led the Series A round of financing with additional investment coming from previous seed investors Amplify Partners, Wing Venture Capital, and NanoDimension Capital. I would like to take this opportunity to welcome Daniel Doctor, Managing Director at Dell Technologies Capital, to the board of directors which includes Mike Dauber, General Partner of Amplify Partners, Jake Flomenberg, General Partner of Wing Venture Capital, Moshe Gavrielov, Independent Director and Former President and CEO of Xilinx, Steven Rosston, Founder and Executive Chairman as well as myself.

The Future — Is your ML Green?™

I am humbled at the opportunity to be a part of such an amazing company with a team of people who are 100% committed to our innovation, execution, and our obsessive focus to help our customers to achieve their goals. However, there is a lot more coming from — so please join us in the movement to reshape what is possible in machine learning at the embedded edge. For more details or to schedule a demonstration under NDA, contact: