Reinforcement Learning / Robot Policy Researcher
Devdoot™ (World Trade Factory™) · DEV 0.0 Humanoid · Remote
About the role
DEV 0.0™ is a humanoid platform built on two-sided honest abstention — the brain won’t act on what it can’t verify, and the body won’t move in ways it can’t make protected (US patent). We’ve built the safety layer. We’re hiring an RL researcher to own the learned control-policy layer beneath it.
What you’ll do
- Train and adapt RL policies for locomotion and manipulation in simulation (MuJoCo / Isaac Lab)
- Run sim-to-real transfer with domain randomization and honest evaluation
- Integrate policies behind our body-agnostic RobotPlatform™ interface, wrapped in our calibrated-confidence / abstention layer
- Benchmark policies across robots and report results as x-of-N, with confident-wrong held at zero at the operating point
What we’re looking for
- Strong RL fundamentals and hands-on policy training (PPO/SAC or similar) in a physics simulator
- Experience with robot control, sim-to-real, or embodied agents
- Comfort with Python + PyTorch, and reading/running open policy checkpoints
- A bias toward measured, reproducible results over demos
Bonus
- Published robotics/RL research, or contributions to open robot-learning stacks (LeRobot, Isaac Lab, ManiSkill, Open X-Embodiment)
- Conformal prediction / uncertainty calibration background
- Humanoid or legged-locomotion experience
Why it’s different Most humanoid roles chase capability. Here your policies are measured on whether they know when they’re wrong — and sit inside a safety architecture designed to make embodied AI trustworthy, not just impressive.
Apply: DM me or start at ? https://devdoot.ai
Simulation-first · research-grade · World Trade Factory™ · Devdoot™ · DEV 0.0™
📌 World Trade Factory - The Future Belongs To You - Reinforcement Learning / Robot Policy Researcher (India)
🏢 SKXYWTF
📍 India