As a Mechanical Engineer focused on Data Collection Hardware, you will bridge the gap between human operators and humanoid kinematics. Your time will be spent in fast-paced rapid development of handheld UMI data-collection tools capable of capturing delicate contact dynamics. Later, there will be the opportunity for rigorous, precision engineering as you architect and build a hand and human exoskeletons.
To train robot foundation models at Persona, we will be relying on massive amounts of high-quality, human-demonstrated data, including rich force and tactile feedback. We are looking for a highly capable Mechanical Engineer to own the physical hardware our operators use to collect data. You will initially be responsible for rapidly iterating on agile, handheld Universal Manipulation Interface (UMI) devices.
What You Will Be Doing
Force-Aware UMI Hardware Design: Design, rapid prototype, and assemble handheld UMI frames, grippers, and mounting rigs that accurately mimic our humanoid’s hand mechanics while integrating load cells and force-torque sensors to capture contact forces.
Sensor & Actuator Integration: Design precise, vibration-resistant mounts for spatial tracking sensors, GoPros, embedded electronics, and force sensors in the UMI. Seamlessly integrate high-resolution encoders and haptic-feedback motors in the console to enable robust, force-aware data collection corresponding with the robot’s senses.
Rapid Prototyping & Manufacturing: Seamlessly shift between rapid in-house 3D printing for UMI iterations and sourcing machined metal components (CNC, sheet metal) for the console structure.
Design for Assembly and Design for Manufacturing: Should be proficient in DFM, DFA, and GD&T concepts.
Defining PCBA outlines when working with EE (mounting points, outline).
Choosing appropriate materials for housing.
Shock-vibe considerations.
Experience with industrial design and the human experience.
Workstation Ergonomics (Human Factors): Apply human-factors engineering to design seating and viewing angles that accommodate the 5th to 95th percentile of human operators during extended, fatigue-free data collection sessions.
