Senior ML Engineer, Manipulation

The Robot Age; San Francisco, Northern (CA, KY); Hybrid; USD 180,000 - 240,000; Full time

Job summary

Chef Robotics is seeking a senior ML engineer focused on manipulation in San Francisco. You will design and train policies for robotic food handling, including imitation learning, reinforcement learning, diffusion policies, and transformer-based action models for real robots.

Join a technically demanding team turning production food-handling data, teleoperation, and evaluation pipelines into manipulation behavior that works in real-world deployments.

Qualifications

• Senior engineering level in robotics, ML, or CS.
• Experience with robot manipulation, visuomotor control, imitation learning, or RL.
• Strong PyTorch and production-minded ML system implementation skills.
• Comfort testing learned policies against real-world robot constraints.

Responsibilities

• Design and train learned manipulation policies for variable food items and end effectors
• Build data collection and training pipelines with teleoperation and kinesthetic teaching inputs
• Evaluate policy architectures such as diffusion policies and action-model transformers
• Adapt models for production robots handling deformable and inconsistent materials

Skills

Robot manipulation

Visuomotor control

Imitation learning

Reinforcement learning

PyTorch

Production ML systems

Education

Robotics, ML, or CS depth

Tools

-

Job description

What the role actually is

Chef Robotics is hiring a senior ML engineer focused on manipulation in San Francisco. The role designs and trains policies for robotic food handling, including imitation learning, reinforcement learning, diffusion policies, and transformer-based action models.

The listing is technical, but it is directly relevant to robot learning: the work turns production food-handling data, teleoperation, and evaluation pipelines into manipulation behavior that can run on real robots.

What you would work on

• Design and train learned manipulation policies for variable food items and end effectors
• Build data collection and training pipelines with teleoperation and kinesthetic teaching inputs
• Evaluate policy architectures such as diffusion policies and action-model transformers
• Adapt models for production robots handling deformable and inconsistent materials

What they are asking for

• Robotics, machine learning, or computer science depth at senior engineering level
• Experience with robot manipulation, visuomotor control, imitation learning, or reinforcement learning
• Strong PyTorch and production-minded ML system implementation skills
• Comfort testing learned policies against real-world robot constraints

Why this one is worth a look

Food manipulation is a hard physical AI problem because the objects are inconsistent, deformable, and messy. Chef's production deployments make this a useful signal for anyone watching robot learning move beyond lab tasks into commercial operations.