Senior Policy Pretraining Engineer - Robotics

Walden Robotics; San Francisco (CA); On-site; USD 210,000 - 320,000; Full time

Benefits offered by this job

Annual bonus

Equity

Insurance

401(k) matching

Flexible PTO

Lunch

Job summary

Walden Robotics is seeking a senior IC to lead policy pretraining, owning core parts of the stack end to end and taking major experiments from idea to trustworthy results. You will set direction others build on and mentor MTS engineers on the stack.

You will design and train policy-pretraining models, run experiments, debug training dynamics, and improve the pretraining pipeline for reliability and scalability.

Qualifications

• Deep hands-on experience training deep models with strong engineering fundamentals in a modern ML framework.
• A track record of driving ambiguous, large-scale training workstreams to a result independently.
• Comfort reasoning about model behavior from data and metrics rather than intuition alone.
• The judgment to make sound modeling and architecture trade-offs and set direction others follow.
• Experience mentoring engineers and raising technical standards.

Responsibilities

• Design and train policy-pretraining models, owning significant parts of the architecture and training loop.
• Design, run, and analyze experiments; debug training dynamics; iterate quickly.
• Build reusable components and drive the efficiency and reliability of the pretraining pipeline.
• Investigate model behavior from data and metrics, and turn findings into concrete improvements.
• Set technical direction for pretraining and mentor MTS engineers on the stack.

Skills

Model Training

Ownership

Data-Driven Reasoning

Technical Judgment

Leadership

Job description

Walden Robotics is seeking a senior IC to lead policy pretraining, owning core parts of the stack end to end and taking major experiments from idea to trustworthy results. You will set direction others build on and mentor MTS engineers on the stack.

You will design and train policy-pretraining models, run experiments, debug training dynamics, and improve the pretraining pipeline for reliability and scalability.