Full-Stack Robotics & AI Engineer

Position Summary

We are looking for an engineer who can take a learning-based robotics system from model development to reliable operation on a physical robot. You will work across perception, policy learning, inference, and robot integration, connecting what a model predicts to what a robot actually does.

This is a hands-on engineering role in our Robotics Lab in Emeryville, California. You will work closely with researchers and hardware engineers to develop manipulation capabilities, investigate failures on real robots, and improve the software and models together.

Core Responsibilities

• Develop models for robotic manipulation: fine-tune and evaluate vision and action models for robotic manipulation, using pretrained models and robot data.
• Connect perception to action: build software connecting cameras and other sensors to model inference and robot commands, including calibration, coordinate transforms, action representations, and control interfaces.
• Optimize models for deployment: deploy models on robot hardware and optimize inference latency, memory use, and throughput through profiling, distillation, compilation, and precision changes.
• Build responsive policy execution, including asynchronous inference and action chunking where appropriate.
• Diagnose failures across the stack: debug failures from observations and model outputs to middleware and physical motion, and build logging, replay, and evaluation tools.
• Build reusable robot software: package robot capabilities with clear interfaces, reproducible configurations, and tests.

Desired Qualifications

• Machine learning engineering: strong Python skills and practical experience training or fine-tuning models in PyTorch or a comparable framework.
• Physical robot integration: experience integrating learning or perception systems with physical robots, including coordinate frames, sensor calibration, and action or control interfaces.
• Systems debugging and optimization: ability to profile software, identify runtime bottlenecks, and debug problems spanning models, robotics software, and hardware.
• Practical engineering experience: a record of building working systems and evaluating them on real tasks.
• Technical collaboration: comfort working closely with researchers and engineers, making practical tradeoffs, and iterating from experimental results.
• Robot learning experience with frameworks such as LeRobot or openpi, imitation learning, or vision-language-action models (helpful, not required).
• Inference optimization experience with tools such as ONNX or TensorRT (helpful, not required).
• Manipulation and policy execution experience, including dexterous or contact-rich manipulation, asynchronous policy execution, or reusable robot skill interfaces (helpful, not required).