Full-Stack Robotics & AI Engineer
Hands-on robotics engineer at Astera Institute's new robotics lab in Emeryville, CA, taking learning-based manipulation systems from model development to reliable operation on physical robots. Expect fine-tuning PyTorch vision/action models, wiring perception to robot control, and optimizing inference for on-robot deployment.
- Compensation
- $180K–$280K / yr
- Location
- Emeryville, United States
- Work setup
- Onsite
- Seniority
- Not specified
- Employment
- Not specified
- Experience
- Not specified
- Category
- Fullstack
- Posted
Skills
Must have
- Strong Python skills and practical experience training or fine-tuning models in PyTorch or a comparable framework
- Experience integrating learning or perception systems with physical robots, including coordinate frames, sensor calibration, and action or control interfaces
- Ability to profile software, identify runtime bottlenecks, and debug problems spanning models, robotics software, and hardware
- A record of building working systems and evaluating them on real tasks
- Comfort working closely with researchers and engineers, making practical tradeoffs, and iterating from experimental results
Nice to have
- Robot learning frameworks such as LeRobot or openpi, imitation learning, or vision-language-action models
- Model distillation and inference optimization with tools such as ONNX or TensorRT
- Dexterous or contact-rich manipulation, asynchronous policy execution, or reusable robot skill interfaces
Full description
About Astera
Astera Institute is a private foundation supporting ambitious work in science and technology through grants, investments, and in-house research. Its research programs combine the pace and autonomy of a startup with the freedom to pursue difficult scientific questions. Learn more about Astera’s mission at .
About the Robotics Lab
We are building a frontier robotics lab at Astera Institute. We believe neuroscience and AI will converge in embodied intelligence: animals and robots face shared challenges of perceiving their surroundings, navigating, manipulating objects, and using memory and reasoning to accomplish tasks in the physical world. Robotics provides a powerful testbed for investigating how biological neural circuits solve these challenges and for developing AI systems that turn perception, learning, and cognition into effective action.
Our strategy starts with making robots useful, while developing a mechanistic understanding of how neural circuits can generate intelligent behavior. We aim to continually expand the repertoire of behaviors a robot can perform reliably in real environments, while improving its ability to learn new skills, adapt to changing conditions, and recover from failure. These practical capabilities provide both measurable outcomes and a foundation for investigating deeper principles of intelligence.
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 the software that connects 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. Implement responsive policy execution, including asynchronous inference and action chunking where appropriate.