San Francisco Bay Area
Learning from data at massive scale is finally unblocking robotics progress. Robots will soon become ubiquitous. Without knowing exactly what they'll be asked to do, or where they'll be asked to go, we want to be able to trust their actions to be safe, responsive, natural, and expressive. This is the problem I'm working on at Sunday.
building robots to give you back your time
• improved arm position accuracy 10x, by combining multiple new methods across signal processing, dynamics modeling, trajectory refinement, and kinematic calibration • implemented sensor fusion for a new SOTA method of detecting & localizing multiple body contacts • built new pipeline for programmatic generation & simulation of low-fidelity robot models, to accelerate and scale evaluation of early-stage design concepts • co-designed new virtual dynamics for teleop system with bilateral force feedback, achieving better robustness to singularities and improved stability while retaining haptic transparency • contributed to new actuator designs and sensing architectures for next-gen robot arms • coordinated w/ 20+ colleagues across 4 teams on 2 continents to deploy new calibration methods, QC tests, firmware upgrades, and control SW changes to robots in mass production
• evaluated performance of 6-DOF F/T sensors through both experiments and sim • integrated external sensors w/ arm, tested inertial parameter estimation methods
• TA for ME 104- Mechatronics: ran lab sessions for juniors/seniors covering the use of microcontrollers, sensors, and actuators in multiple robotics projects • TA for ME 127- Intro to Mechanisms: provided students with support for circuit design and Arduino programming during a quarter-long electromechanical design project
• Integrated power distribution and control systems for a rocket landing platform vessel for a private space technology company • Designed & ran 6-month project to lay groundwork for a new companywide data strategy, wrote 50 pg report, shared findings w/ leadership and at all-hands mtg, inspired follow-on investment • Developed new workflows using EPLAN to efficiently manage large-scale cable block diagrams • Automated data processing tasks w/ VBA and AutoLISP for 3 projects, saving hundreds of hours