Robot Learning Engineer - Manipulation
$60k - $300k • Sunnyvale • FullTime
Posted 8h ago
Remote Work Policy
On-site
Employment Type
FullTime
Categories
Applied AI Engineer
About the job
Applied Intuition is building a robot learning platform on Dana, its physical AI platform, designed to enable any robot to learn and continuously improve industrial tasks. The robotics team develops this platform and applies it to deliver robot autonomy for customers, involving the training, evaluation, and deployment of policies on real robots performing industrial tasks. Team members engage directly with hardware and witness their work's impact on customers. As a Robot Learning Engineer, you will be responsible for manipulation policies from their initial definition through to reliable execution on a physical robot. Your role will involve training and fine-tuning these policies, diagnosing failures, and ensuring their dependability for customer use, with success measured by performance on actual robots rather than solely offline benchmarks. The company values practical experience with real systems over specific degrees or titles and is open to candidates at various experience levels.
Responsibilities
- Work on the full learning loop for manipulation tasks, including task definition, demonstration collection, data curation, training, real-robot evaluation, and deployment.
- Train and fine-tune manipulation policies, selecting appropriate approaches from large pretrained models like vision-language-action models to compact task-specific policies.
- Develop repeatable methods for industrial tasks such as pick-and-place, bimanual handling, and contact-rich assembly, incorporating force or tactile signals as beneficial.
- Deploy policies on edge compute systems and validate observation processing, action interfaces, and control timing on the robot.
- Improve data and models by analyzing failures and human interventions.
- Measure and enhance key customer-focused metrics like success rate, cycle time, and intervention rate.
- Package learned recipes and models to facilitate the application of knowledge to new tasks and robots.
Requirements
- Trained or fine-tuned a learned manipulation policy and deployed/evaluated it on a physical robot.
- Strong Python and PyTorch skills for writing maintainable training, evaluation, and deployment code.
- Practical depth in imitation learning and at least one modern policy family (e.g., vision-language-action models, diffusion policies, action-chunking transformers).
- Working knowledge of robot kinematics, coordinate frames, camera calibration, and the interface between learned actions and low-level control.
- Habit of diagnosing failures through controlled experiments involving data, sensing, model, and execution.
- Comfort with open-ended problems and clear communication of tradeoffs to teammates.
Benefits
- Base salary
- Equity (options and/or restricted stock units)
- Comprehensive health, dental, vision, life and disability insurance coverage
- 401k retirement benefits with employer match
- Learning and wellness stipends
- Paid time off