Robotics
Tesla Trade Suit Settled: Optimus Engineer Launches Proception with $11M to Perfect Robot Hands
After resolving a legal dispute with his former employer, Jay Li is building specialized robotic hands trained on human motion data captured through sensor-laden gloves.
TechCrunch — AI
Jay Li, a former Tesla engineer who led work on the Optimus humanoid robot program, has moved past a trade secret lawsuit with the electric-vehicle maker and is now pursuing what he sees as an even tougher engineering challenge: giving robots hands that work like humans.
Li's startup, Proception, announced a $11 million seed funding round on Monday, led by First Round Capital and including backing from Y Combinator and BoxGroup. The company is simultaneously beginning to ship its first dexterous robot hands to research groups and robotics firms while opening orders to a broader market.
Resolving the Legal Battle#
Tesla sued Li last year for allegedly taking proprietary information about robotic manipulation when he left the company to launch Proception. The legal dispute stretched on for months, but a settlement reached earlier this month ended the case, clearing the way for him to move forward without encumbrance.
When asked about the experience, Li framed it as a test of mettle. "I think it's kind of like a resilience test," he noted, adding that surviving the challenge has left him and his team stronger.
The Hand Problem#
The robotics industry has poured vast resources into humanoid platforms, yet most experts agree that creating robot hands capable of true dexterity—performing intricate manipulation tasks—remains one of the field's hardest unsolved problems. Even Elon Musk has publicly identified advanced hand control as a major engineering hurdle.
Kevin Lynch, director of Northwestern University's Center for Robotics and Biosystems, previously told the Wall Street Journal that functional, human-equivalent robot hands are likely a decade away. Li believes Proception can compress that timeline significantly through a fundamentally different approach to training.
A New Data Collection Method#
Most robotics companies train their systems using teleoperators—people wearing VR headsets who guide robots remotely, with the systems learning from those commands. However, this method lacks tactile feedback for the operator and scales only to the number of physical robots a company possesses.
Proception's solution centers on a sensor-studded glove that humans wear during training sessions. This captures rich motion and interaction data without requiring a robot present, enabling faster, more scalable data collection. The same glove technology is integrated into Proception's hand itself, serving as its sensory "skin."
