Staff ML Researcher: Planning & Interventional Causality

Kindredventures; San Francisco (CA); On-site; USD 150,000 - 230,000; Part time

Job summary

Kindredventures in San Francisco seeks researchers to build the planning layer on top of a Large Physics foundation Model (LPM), enabling objectives conditioning and producing actions to affect outcomes from operational decisions to physical interventions.

We value a relentless problem-solving approach and rapid execution. You will work across data, model, eval, and infrastructure, taking ideas from prototype to scaled training runs with a focus on uncertainty and decision quality.

Qualifications

• Strong ML fundamentals with depth in RL or planning.
• Experience training models and analyzing experimental results.
• Familiarity with reasoning, planning, or acting with learned models.
• Track record turning open-ended research into working systems.

Responsibilities

• Research and implement methods that turn a predictive physics model into one that reasons toward objectives.
• Develop approaches for decision-making under uncertainty in high-dimensional spaces.
• Build interfaces for specifying objectives and constraints, and actions to satisfy them.
• Run experiments and ablations to connect reasoning to decision quality.
• Work across the full ML stack—from data to infrastructure—scaling training runs.

Skills

ML fundamentals

Reinforcement learning

Planning and control

Decision-making under uncertainty

Model-based RL

Post-training of large models

Ablation studies

Experiment analysis

Job description

Kindredventures in San Francisco seeks researchers to build the planning layer on top of a Large Physics foundation Model (LPM), enabling objectives conditioning and producing actions to affect outcomes from operational decisions to physical interventions.

We value a relentless problem-solving approach and rapid execution. You will work across data, model, eval, and infrastructure, taking ideas from prototype to scaled training runs with a focus on uncertainty and decision quality.