In-depth profile of Refraction AIin Ann Arbor's tech ecosystem
Refraction AI is a small company with an oversized role in the future of last-mile delivery. Founded in 2019 by Matthew Johnson-Roberson and Ram Vasudevan, both U-M faculty, Refraction builds autonomous robotic delivery vehicles that are small enough to fit in a bike lane and cheap enough to deploy at scale. The company's REV-1 robots have been operating in Ann Arbor for years, delivering groceries and restaurant orders across the city.
The engineering bet is deliberate. Rather than chasing full autonomy in standard automobile form factors, Refraction targeted a segment that sidesteps most of the catastrophic failure modes. A delivery robot that travels in bike lanes, weighs a couple hundred pounds, and tops out at fifteen miles per hour is a fundamentally different safety case than a three-thousand-pound passenger vehicle on a highway. That design choice has let the company deploy real services in real cities while the heavier AV programs have spent years in simulation.
Refraction's Ann Arbor headquarters puts the company close to U-M's robotics program, Mcity, and the local operating environment where the REV-1 has logged the most miles. The city itself has become a proving ground for the robots, with residents ordering deliveries in the regular course of business. That real-world operating data is an asset that almost no other delivery robotics company has at this depth.
For engineers interested in applied robotics, computer vision, and cyber-physical systems that actually ship to customers, Refraction offers a compact, high-agency environment. The office is central to Ann Arbor, the team is small enough that every engineer ships impactful work, and the cost of living lets a robotics engineer live comfortably on a startup salary.