Senior Research Scientist / Engineer: Robot Learning

We are looking for a Senior Research Scientist / Engineer to lead policy training and real-world post-training for FRIDAY, our mobile humanoid robot. You will develop robust learned behaviors for dexterous and whole-body mobile manipulation by combining training in our in-house simulator, Holiday-Newton, with demonstrations and real-robot experience. Working closely with researchers and engineers across dynamics, planning, control, and simulation, you will translate advances in reinforcement learning and imitation learning into robust, dexterous behavior on the physical robot., * Set technical direction and own FRIDAY’s policy-training and post-training stack, including task formulation, observations, action spaces, rewards, curricula, and evaluation protocols.

• Train and compare policies at scale in Holiday-Newton using calibrated digital twins, optimizing for sample efficiency, policy performance, and real-time execution; validate robustness through Sim-to-Sim testing before deployment.
• Deploy, evaluate, and debug policies on FRIDAY, and collect real-world visual and tactile data, demonstrations, interventions, rollouts, and failure cases.
• Develop real-world fine-tuning and real-robot policy-adaptation methods, feeding physical results back into datasets, task definitions, and training procedures.
• Collaborate with dynamics, planning, control, and simulation researchers and engineers on digital twins, system identification, and Sim-to-Real transfer, and diagnose whether failures arise from policies, data coverage, model mismatch, or physical execution.

• BS, MS, or Ph.D. in Robotics, Computer Science, Mechanical Engineering, Electrical Engineering, or a closely related quantitative field.
• 7+ years of robotics experience, or 5+ years with an advanced degree, focused on training, evaluating, and deploying learned policies for physical robotic systems.
• Strong foundation in robot learning, with depth in reinforcement learning, imitation learning, policy optimization, or related approaches.
• Hands-on experience with GPU-accelerated policy training at scale using Python and JAX or PyTorch, together with simulators such as MuJoCo, Warp, Drake, or comparable platforms; working proficiency in C++.
• Experience with Sim-to-Sim policy validation across physics engines, simulator configurations, or model variants.
• Experience deploying policies on physical robots and improving them using demonstrations, interventions, autonomous rollouts, or failure data through real-world fine-tuning or policy adaptation., * Hands-on experience with tactile sensors, including collecting and using real-world visual and tactile sensor data to train and evaluate robot-learning policies.
• Familiarity with modern 3D reconstruction and scene-representation methods-including Gaussian Splatting-and their use in synthetic-data generation, Real2Sim asset creation, and digital-twin workflows.
• Familiarity with Sim-to-Real transfer, including calibrated digital twins, system identification, dynamics randomization, actuator modeling, latency modeling, and sensor-noise modeling.

• Compensation: $200,000-$300,000 USD base salary + competitive equity. Final compensation is based on experience and technical impact.
• Medical, dental & vision: 100% employer-paid coverage for employees and partial coverage for dependents. Medical coverage is provided through a full-network PPO plan.
• 401(k): Employer-sponsored retirement plan.
• Relocation: Stipend for out-of-state candidates.
• Meals: Free daily meals.

Holiday Robotics Research Inc. is the U.S. subsidiary of Holiday Robotics, a leading humanoid robotics company based in South Korea. We build the entire robotics stack in-house-from hardware and firmware to planning, control, simulation, and end-user software-with the goal of freeing people from tedious, dangerous, and repetitive tasks.