Member of Technical Staff, VLA

About DeepReach

DeepReach is building the next-generation data infrastructure for robotics. We help bridge the gap between promising robot models and real-world deployment by building the systems, data pipelines, and learning loops needed to make robots improve in production.

We believe robotics progress will be driven not just by better models, but by better data engines: how data is collected, filtered, evaluated, and turned into measurable gains on real tasks. Our team works across robot deployment, teleoperation, data generation, model training, and evaluation, with a strong bias toward hands-on execution and fast iteration.

The Role

As a Member of Technical Staff, VLA, you will develop robot learning systems that connect data, models, and deployment. You will work on training and adapting VLA and policy models for real-world robotic tasks, using data collected from teleoperation and real operations rather than relying only on static public datasets.

This is a highly applied role for someone who can read papers, reproduce results, and then push beyond them in the messy reality of physical systems. You should be comfortable moving between PyTorch training runs, evaluation design, deployment debugging, and hands-on work with robot hardware when needed.

What You’ll Do

• Train and fine-tune VLA, diffusion policy, or related robot learning models for real-world tasks
• Build data and training pipelines that turn deployment and teleoperation data into better policy performance
• Design experiments to identify what actually improves task success rates on real robots
• Work closely with the data and deployment teams to close the loop between model failures and data collection strategy
• Deploy and debug learned policies on physical robot systems, including robot arms, grippers, and multi-camera setups
• Define internal evaluation frameworks tied to real operational tasks rather than benchmark-only performance
• Read and implement recent papers, reproduce promising results, and adapt them for our stack and constraints

What We’re Looking For

• Bachelor’s degree or equivalent practical experience in Robotics, Computer Science, Electrical Engineering, or a related field
• Strong hands-on background in robot learning, including one or more of: imitation learning, RL, diffusion policies, VLA, or visuomotor policies
• Strong PyTorch skills and experience with modern model training workflows
• Ability to move from paper to implementation quickly and independently
• Strong systems intuition across perception, policy, and control
• Real-world experience deploying or debugging policies on physical robots
• High ownership mindset and comfort operating in an early-stage, fast-changing environment

Strong Signals

• You have trained and deployed learned policies on real robot systems
• You have worked with noisy, imperfect deployment data and still shipped improvements
• You have built your own evaluation or experimentation framework instead of relying only on standard benchmarks
• You are excited by the gap between “works in the paper” and “works in production”
• You like being close to hardware and can debug end-to-end failures across software and robot systems

Why Join Us

• Work on real robotics problems where success is measured by deployment performance
• Help define how data scaling works for embodied intelligence in production environments
• Build core learning systems with direct influence on company direction and technical roadmap
• Join an onsite team that values speed, ownership, and technical depth
• Benefits include health insurance, free food, 401(k), and generous PTO

Location

Bay Area, CA (Onsite)

How to Apply

Click Apply Now and submit your resume along with links to relevant projects, publications, GitHub, or portfolio work.

We run a fast interview process and aim to move quickly for strong candidates.

If you encounter any issues while applying, please contact [email protected] with a screenshot.