Research Scientist, Reinforcement Learning

Pramaana Labs•Palo Alto, CA

About The Position

Pramaana Labs is a frontier AI lab building a verification layer for AI. Founded in 2025 and headquartered in Palo Alto, the company turns complex human knowledge—including tax codes, legal rules, clinical guidelines, and security protocols—into formal representations so AI answers can be traced, challenged, and proved. Its architecture pairs foundation models trained to formalize and reason with a symbolic world model encoded in Lean 4. The company was founded by a team from Google, DeepMind, and Glean, combining frontier AI researchers and formal methods experts. In this role, the researcher will help train foundation models to interact natively with symbolic world models and scale their reasoning capabilities. The work applies reinforcement learning with verifiable rewards (RLVR) to new domains, uses rewards verified by the Lean kernel, advances autoformalization and theorem proving, and develops RL algorithms, data, and evaluations. The researcher will also define how models use test-time compute to solve long-horizon logical tasks, owning the full loop from rollout sampling and reward design through model updates and evaluation.

Requirements

  • Deep, hands-on research experience applying reinforcement learning to reasoning models at scale.
  • Experience with RLVR, test-time RL, or exact deterministic reward signals.
  • Strong algorithmic and systems intuition.
  • Comfort writing custom RL loops, managing data pipelines, and building robust evaluations from scratch.
  • High autonomy and ability to take a loosely defined problem area and independently drive it to state-of-the-art results.

Nice To Haves

  • Background in formal verification, such as Lean, Coq, or Isabelle.
  • Strong publication record in top-tier venues such as NeurIPS, ICML, or ICLR.

Responsibilities

  • Scale RLVR to new real-world domains using the Lean kernel as a deterministic reward signal.
  • Design and implement novel RL algorithms and test-time compute optimizations tailored to formal environments and proof search.
  • Advance autoformalization by training models to map highly technical natural language into strict formal specifications.
  • Curate data and build evaluations that closely correlate with verifiable downstream reasoning capabilities.
  • Own the full research loop, including rollout sampling, reward design, model updates, and evaluation.
© 2026 Teal Labs, Inc
Privacy PolicyTerms of Service