SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Applied Intuition is building a robot learning platform (Dana) that provides the data infrastructure and training intelligence for companies to make robots learn industrial tasks and continuously improve. The robotics team works hands-on with hardware, seeing their work deployed on real robots performing real industrial tasks.
As a Robot Learning Engineer specializing in manipulation, you will own the full learning loop for manipulation tasks: from task definition and demonstration collection through data curation, training, real-robot evaluation, and deployment. You will train and fine-tune manipulation policies using approaches ranging from large pretrained vision-language-action models to compact task-specific policies, selecting the right approach for each task. You'll develop repeatable recipes for industrial tasks such as pick-and-place, bimanual handling, and contact-rich assembly, incorporating force or tactile signals where beneficial.
You will deploy policies on edge compute, validate observation processing, action interfaces, and control timing on physical robots. You'll turn failures and human interventions into better data, models, and evaluation strategies. Success is measured on the robot, not only on offline benchmarks. You'll measure what customers care about—success rate, cycle time, intervention rate, and the data and time needed to reach targets—and improve those metrics task by task. Finally, you'll package recipes and models so subsequent tasks and robots benefit from prior learning.
Applied Intuition is a $15 billion Series A company founded in 2017, headquartered in Sunnyvale with offices globally. It powers physical AI across automotive, defense, trucking, construction, mining, and agriculture. Eighteen of the top 20 global automakers and the U.S. military trust the company's solutions.
The company is in-office, expecting full-time employees to work from the Sunnyvale office 5 days per week, though occasional flexibility is recognized for morning remote work or family commitments.
REQUIREMENTS:
- Trained or fine-tuned a learned manipulation policy and deployed and evaluated it on a physical robot
- Strong Python and PyTorch skills with ability to write maintainable training, evaluation, and deployment code
- Practical depth in imitation learning and at least one modern policy family (vision-language-action models, diffusion policies, or 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 with controlled experiments across data, sensing, model, and execution
- Comfort taking on open-ended problems and communicating tradeoffs clearly to teammates at the robot
NICE TO HAVE:
- Experience with bimanual manipulation, force-controlled insertion, tactile sensing, or dexterous hands
- Experience with ROS 2, LeRobot, or simulators such as Isaac and MuJoCo
- Experience optimizing inference on NVIDIA Jetson or GPU edge systems (model export, compilation, quantization)
- Experience with reinforcement learning post-training, learning from interventions, or transferring policies across robot platforms
