Lacksap
2026A custom 10 actuator biped built with QDD motors.
Theo Moore-Calters
I’m a research engineer working in robotics and simulation. My work focuses on reinforcement learning for robot locomotion and extends into loco-manipulation, sim-to-real transfer, SLAM, and state estimation.
Currently working on
Selected work
Duck Mini Pro Headless
2026A custom 10 actuator biped built with servo motors. It was trained using reinforcement learning in MJLab and deployed using ROS 2 on a Raspberry Pi 5.
Velocity Hub
2026A suite of velocity-tracking environments for bipeds and quadrupeds across MJLab and MuJoCo Playground. It provides reusable robot configurations, pretrained policies, and an interactive browser demo for comparing learned locomotion across platforms.
Sports Hub
2025A collection of reinforcement-learning environments for humanoid and quadruped locomotion and loco-manipulation. Policies trained with JAX and Brax can be run directly in the browser through an interactive MuJoCo demo.
Industry experience
Luda
Research Engineer
Jan 2025 – Jan 2026Research Scientist Intern
Jul 2024 – Jan 2025Developed reinforcement-learning environments and policies in MJX for locomotion and loco-manipulation across humanoid and quadruped robots. Built competitive, locomotion-based multi-agent tasks, including mixed-morphology match-ups, and engineered a parallelised IPPO training framework.
SECCL
Software Engineer Intern
Built backend and React features for a custody and investment-technology startup. Developed an API workflow that consolidated clients’ fractional shares for daily execution on the London Stock Exchange, and implemented ISA and JISA support.