Elementary, my dear Robot
By Mo Islam, Partner, Threshold Ventures
I’m thrilled to announce Threshold is leading a $12.7 million Series A investment in Elementary Robotics, a software-defined robotics company applying deep learning-based computer vision to industrial inspection. Elementary uses proprietary software and low-complexity, commodity hardware to automate quality and traceability workflows in manufacturing and logistics. I’ve joined the board, and am working alongside a great group of seed investors from Fika, Fathom, Idealab, Ubiquity, and Toyota.
I met Elementary Co-Founder and CEO, Arye Barnehama, a year ago. We started regularly getting together for dinner to discuss our shared interest in AI and robotics. From the beginning, I could see Arye has an incredible product intuition. We quickly realized that we held a strong shared belief in the promise of computer vision, the killer application of AI. We both admire what machines have learned to do incredibly well because of the deep learning revolution. Today he has taken that vision and found a huge customer problem to solve around visual inspection in the trillion-dollar manufacturing industry.
I’ve been excited about deep learning applications in computer vision for many years now, and it’s helped inform prior investments in medical imaging diagnostics and autonomous vehicles. Deep learning is the state-of-the-art in computer vision algorithms today. The superhuman performance that machine learning models can achieve with large amounts of high-dimensional data (like imagery, you say?) is an incredible enabling technology.
We’ve already hit this threshold in computer vision performance on a variety of visual learning tasks. We’re on an exponential growth curve of the rapid adoption of new applications enabled by computer vision. In manufacturing, computer vision has huge potential.
In the wake of Industry 4.0, transforming traditional manufacturing to intelligent manufacturing, companies are looking for productivity gains in throughput and lower operating costs using automation. The industrial machine vision market in particular is expected to reach $13 billion over the next few years, with quality assurance and inspection as the largest market segment. Cognex and Keyence are sleepy incumbents with enviable operating profits, but neither are well-equipped to innovate in software, the source of the largest value creation. I’m convinced Elementary will lead the next-generation of industrial machine vision companies with a combination of beautiful product design, rapidly deployable hardware, and deep learning-based software.
From first principles, robots are defined as anything that senses, computes and actuates. Robotics has been at the forefront of many historical investments at the former DFJ Venture and continues to be an active thesis area for investment at Threshold today. However, I’m convinced the next generation of robotics companies will be completely software-driven. These companies will think software-first, and leverage low-complexity and commodity hardware to deploy their intelligence. These robotics companies will have a first-principles mindset to build a rapidly deployable product that solves the customer’s production problem, not design research projects with an expensive bill of materials (BOM) stuck in endless pilots. Elementary has already accomplished this and more with their work with Toyota at their Indiana manufacturing plant and has deployed in production with many other undisclosed customers.
When it comes to solving problems for humans, I believe robots have to work collaboratively with humans. In manufacturing, this is best served with a man-machine-fusion approach. There’s a strong foundation for continuous learning by using the domain experience and intuition of human operators on the ground to provide feedback to a global AI model. Elementary’s software was built with this approach in mind, providing a data dashboard for operators on the manufacturing line who can continuously improve their AI.
When I went onsite at Elementary and Arye showed me their lab, it was clear he was building a next-generation robotics company. He built a software-first, product-led sales organization. He and his CTO Dan Pipe-Mazo built an amazing team of hardware, UX, and machine learning engineers. They deployed in production with large electronics manufacturers and automotive companies and were obsessed with building a rapidly deployable, highly intuitive product to improve manufacturing productivity.
I spoke to a number of inspectors that worked with the Elementary system. It was exactly the man-machine symbiotic relationship Arye and I discussed at several of our dinners: offloading repetitive, low-cognitive-order tasks to the machine to increase productivity, with humans expanding to more complex tasks was something they were ecstatic about. It wasn’t about automating away people but expanding their horizons.
COVID-19 is having a tremendous impact on manufacturing globally. The economic crisis has accelerated demand for products that can lower cost, provide remote monitoring capabilities, enhance employee safety, and literally reduce the number of humans needing to touch things. Elementary’s solution has provided ways for customers to maintain product quality protocols and reduce costs at a crucial time, which is why we’ve seen demand accelerate during COVID.
I’m privileged to partner with Arye and the Elementary team, and I look forward to working closely with them to help their customers achieve value across the production line, enhance worker safety, and build the future of manufacturing.
