Staff Software Engineer, Learned & Hybrid Behavior Planning

Lead Planning-side integration of learned models into the behavior planning stack.

Collaborate closely with ML teams on model improvements, requirements, evaluation, and deployment.

Work on learned planning components as well as other ML-driven planning signals, such as behavior classification, actor intent understanding, and data-driven decision-making.

Design integration strategies that balance learned components with existing heuristic planning systems.

Define validation, fallback, monitoring, and safety criteria for learned planning components.

Debug and analyze model behavior using simulation, logs, metrics, and real-world autonomy data.

Partner with cross-functional teams across Perception, ML, Planning, Simulation, Systems, and Safety.

Lead technical designs and mentor other engineers.

Strong experience in autonomous vehicles, robotics, or a related autonomy domain.

Deep technical background in behavior planning, decision-making, or motion planning.

Strong software engineering skills with proficiency in C++. Python proficiency is a plus.

Experience working with heuristic or classical planning systems.

Experience integrating or developing learned behavior policies, behavior classification, trajectory prediction, or actor intent models.

Ability to reason about safety, system behavior, evaluation, and deployment risk.

Excellent cross-functional communication and technical leadership skills.

The pay range listed below reflects the base salary in our SF/Silicon Valley location, across several internal levels. Actual starting pay will be based on job-related factors including: work location, experience, relevant training, education, skill level and performance during interview. Total compensation at Kodiak includes base pay, equity, bonus and a competitive benefits package