A new startup, Sunday Robotics, has exited stealth mode with $35 million in funding and recruited several senior engineers from Tesla's AI and robotics teams. The company is developing a wheeled household robot called Memo, differing from Tesla's bipedal Optimus. This talent drain highlights intensifying competition in robotics.
Sunday Robotics, founded by Stanford roboticists Tony Zhao and Cheng Chi—Zhao previously interned at Tesla Autopilot—announced on November 24, 2025, that it has raised $35 million in funding led by Benchmark and Conviction. The startup has attracted key talent from Tesla, including Nishant Desai, a nearly five-year veteran on Tesla's machine learning team for Autopilot and Full Self-Driving (FSD); Nadeesha Amarasinghe, former Engineering Lead for AI Infrastructure responsible for backend systems training FSD and Optimus, with over seven years at Tesla; and Perry Jia, who spent almost six years at Tesla leading data engine programs for Optimus and Autopilot and now leads Data Operations at Sunday. Additionally, Jason Peterson, a talent recruiter for Tesla's Optimus and Robotaxi programs, left Tesla in September to join Sunday.
Unlike Tesla's general-purpose humanoid Optimus, which is bipedal, Sunday Robotics' debut robot, Memo, is a wheeled domestic robot designed for household chores such as cleaning dishes and folding laundry. By forgoing legs, Sunday focuses on dexterity and reliability. Memo is trained on a dataset of 10 million behavioral episodes, which the company describes as providing a “ChatGPT moment” for physical movement.
Sunday's data collection method diverges from Tesla's approach. While Tesla relies on VR teleoperation suits—where operators wear motion-capture suits to mimic tasks in labs, a process described as high-fidelity but slow and expensive—and now claims to train on video, Sunday uses a $200 'Skill Capture Glove.' These gloves were distributed to hundreds of ordinary people, termed “Memory Developers,” who recorded themselves performing chores in their own homes. This crowdsourcing enabled Sunday to gather 10 million episodes of real-world data, including messy kitchens, varied lighting, and interruptions like cats on counters, at a lower cost than Tesla's teleoperation labs. The gloves align with Memo's simpler hands, potentially making them more reliable and affordable.