Robot Learning Engineer - Manipulation

Applied Intuition; Sunnyvale (CA); On-site; USD 140,000 - 190,000; Full time

Job summary

Applied Intuition is hiring a robot learning engineer to work on manipulation for its Dana platform. You will own the full learning loop from task definition to data curation, training, evaluation, and deployment on real robots.

The role emphasizes real-robot success and industrial relevance, not just offline benchmarks, with a practical path to deployed autonomy across tasks.

Qualifications

• Experience training or fine-tuning learned manipulation policies.
• Depth in imitation learning and modern policy families (e.g., diffusion policies, action-chunking transformers).
• Strong Python and PyTorch skills for training, evaluation, and deployment code.

Responsibilities

• Train and fine-tune manipulation policies for industrial robot tasks.
• Build repeatable recipes for pick-and-place and bimanual handling.
• Validate policy behavior on physical robots with observation processing and timing.
• Turn failures into better data, models, and evaluation loops.
• Package learned behaviors so new tasks/ platforms can start from prior work.

Skills

Imitation learning

Policy learning

Python programming

PyTorch

Robot manipulation

Experiment design

Tools

Python

PyTorch

Camera calibration

Job description

What the role actually is

Applied Intuition is hiring a robot learning engineer for manipulation work on Dana, its physical AI platform. The role covers the full learning loop: task definition, demonstration collection, data curation, training, evaluation, and deployment on physical robots.

The listing is explicit that success is measured on real robots in customer-relevant industrial tasks, not just offline benchmarks. That makes it a practical robot-learning role with a direct path to deployed autonomy.

What you would work on

• Train and fine-tune manipulation policies for industrial robot tasks
• Build repeatable recipes for pick-and-place, bimanual handling, and contact-rich assembly
• Validate policy behavior on physical robots, including observation processing and control timing
• Turn failures and human interventions into better data, models, and evaluation loops
• Package learned behaviors so new tasks and robot platforms can start from prior work

What they are asking for

• Experience training or fine-tuning learned manipulation policies
• Practical depth in imitation learning and modern policy families such as VLA models, diffusion policies, or action-chunking transformers
• Strong Python and PyTorch skills for training, evaluation, and deployment code
• Working knowledge of robot kinematics, coordinate frames, camera calibration, and low-level control interfaces
• Comfort diagnosing failures through controlled experiments across data, sensing, model, and execution

Why this one is worth a look

Manipulation is where robot learning has to prove it can survive messy, physical work. This role is interesting because it sits inside a platform company trying to make learned industrial robot behavior repeatable across tasks, not just impressive in one demo.