Pramaana Labs
Research Scientist, Reinforcement Learning
Palo Alto · On-site · Full-time · Posted Aug 24, 2026
About the job
Develops and evaluates reinforcement-learning methods that train foundation models to reason with Lean-based symbolic world models and solve formal reasoning tasks.
What you'll do
- Scale RLVR to real-world domains using the Lean kernel for verified rewards; design RL algorithms and test-time compute optimizations for proof search.
- Train models to convert technical natural language into formal specifications, and curate data and build reasoning evaluations.
- Own the training loop, including rollout sampling, reward design, model updates, and evaluation.
What they ask for
- Hands-on research experience applying reinforcement learning to reasoning models at scale.
- Experience with RLVR, test-time RL, or exact deterministic reward signals.
- Algorithmic and systems intuition, including writing custom RL loops, managing data pipelines, and building evaluations.
- Ability to independently drive ambiguous research problems toward state-of-the-art results.
Nice to have
- Background in formal verification, such as Lean, Coq, or Isabelle.
- Publication record in top-tier venues such as NeurIPS, ICML, or ICLR.
Tools
Lean 4, Lean, Coq, Isabelle
AI focus
Post-trains foundation models with reinforcement learning for formal reasoning, autoformalization, and long-horizon logical tasks; develops test-time compute methods and evaluations.
Model-written summary · the original posting is authoritative