Robot Learning Researcher/Engineer (Imitation Learning, Foundation Models, RL)

Proception Inc.; Palo Alto (CA); On-site; USD 140,000 - 210,000; Full time

Job summary

Proception Inc. in Palo Alto is seeking a Robot Learning Researcher/Engineer to advance imitation learning, foundation models, and reinforcement learning for general-purpose control policies.

You will design scalable training pipelines and build the data engine behind the models for large-scale, multimodal tasks. You will train multimodal policies and push them to state-of-the-art performance on real robots, collaborating with AI, hardware and perception teams to deliver production-quality

Qualifications

• MS or PhD in Robotics, Computer Science, ML or related field (or equivalent).
• Proven track record training neural networks end-to-end with reproducible results.
• Experience developing robot learning policies (diffusion, 3D, V-L-A, or video action).
• Experience training models at scale with multi-GPU/multi-node setups and large datasets.
• Strong software engineering in Python and PyTorch or JAX; Linux environment; C++ a plus.
• Experience building large-scale data pipelines for robot learning (demo collection, curation, filtering).
• Comfort with high-dimensional actions and multimodal observations.
• Rigorous evaluation: benchmarks, ablations, robust conclusions.
• Hands-on robotics experience with hardware bring-up, teleoperation, data collection.
• Experience with high-performance simulation (MuJoCo, Isaac Gym/Lab) and sim2real.
• Familiarity with contact-rich/dexterous manipulation, tactile sensing, or differentiable sim

Responsibilities

• Design scalable training pipelines for general-purpose manipulation policies.
• Train large multimodal policies and push them to state-of-the-art on real tasks.
• Build data engine behind the models: demonstration collection, curation and dataset design at scale.
• Integrate visual, proprioceptive, and tactile feedback into policy architectures.
• Take policies from training runs to real hardware with on-robot deployment.
• Collaborate across AI, hardware and perception teams to build closed-loop systems.
• Publish or contribute to research while delivering production-grade control stacks.

Skills

Imitation Learning

Foundation Models

Reinforcement Learning

Multimodal Models

Diffusion Policies

3D Policies

Video Action Models

Large-scale Training

Python

PyTorch

JAX

Education

MS or PhD in Robotics, Computer Science, Machine Learning, or related field

Tools

Python

PyTorch

JAX

C++

Job description

Robot Learning Researcher/Engineer (Imitation Learning, Foundation Models, RL)

Join our team to push the frontier of robot learning. You'll train general-purpose control policies at scale - spanning imitation learning and large multimodal models - build the data and evaluation pipelines that make them work, and take policies from training runs to real robots. We care more about your ability to train models that work than about any particular robot, task, or sensor you've used before.

Requirements

• 01 MS or PhD in Robotics, Computer Science, Machine Learning, or related field-or equivalent experience
• 02 Strong track record training neural networks end-to-end: you can take a model from idea to a working, debugged, reproducible result
• 03 Experience developing robot learning policies like diffusion policies, 3D policies, vision-language-action models, or video action models
• 04 Experience training models at scale: multi-GPU/multi-node training, large datasets, long runs, and debugging throughput, stability, and scaling behavior
• 05 Strong software engineering skills in Python and PyTorch or JAX in Linux environments (C++ a plus)
• 06 Experience building large-scale data pipelines for robot learning - demonstration collection, curation, filtering, and dataset design
• 07 Comfortable training policies with high-dimensional action spaces and multimodal observations (vision, proprioception, language)
• 08 Rigorous about evaluation: designing benchmarks, running ablations, and drawing correct conclusions from noisy real-world results
• 09 (+) Hands-on robotics experience - hardware bring-up, teleoperation and real-world data collection, on-robot deployment
• 10 (+) Experience with high-performance simulation (MuJoCo, Isaac Gym/Lab) and sim2real techniques (domain randomization, dynamics adaptation, residual policy learning)
• 11 (+) Familiarity with contact-rich or dexterous manipulation, tactile sensing, or differentiable simulation

Details & responsibilities

• 01 Design and implement scalable training pipelines for general-purpose manipulation policies
• 02 Train large multimodal policies - diffusion, 3D, vision-language-action, and video action models - and push them to state-of-the-the advantage on real tasks
• 03 Build the data engine behind the models: demonstration collection, curation, filtering, and dataset design at scale
• 04 Integrate visual, proprioceptive, and tactile feedback into policy architectures
• 05 Take policies from training runs to real hardware, closing the loop with on-robot deployment and iteration
• 06 Collaborate across AI, hardware, and perception teams to build closed-loop manipulation systems
• 07 Publish or contribute to cutting-edge research while delivering production-quality control stacks
• 01 Competitive salary and meaningful equity
• 02 Full health, dental, and vision insurance
• 03 Access to custom-built dexterous robots
• 04 Collaboration with leading researchers in robotics and AI
• 05 Backed by YC and top-tier investors
• 06 High-ownership role with the opportunity to lead core initiatives in real-world robot learning