Research Scientist, AI

Pramaana Labs; Palo Alto (CA); Sur place; USD 180 000 - 260 000; Plein temps

Résumé du poste

Pramaana Labs, headquartered in Palo Alto, is building the verification layer for AI. We train models to formalize complex knowledge and reason over symbolic world models, focusing on accuracy and reliability across real-world domains.

You will lead ambitious research at the intersection of foundation models, formal reasoning, and verification, shaping production capabilities and measurable improvements in model performance.

Qualifications

• Track record of high-impact research with deployment to product or core tech.
• Deep hands-on experience in foundation model training and post-training workflows.
• Strong background in modern AI research including large-scale models and reasoning.

Responsabilités

• Lead ambitious research at the intersection of foundation models, formal reasoning, and verification.
• Develop novel training and post-training approaches for translating domain knowledge into structured representations for verifiable reasoning.
• Advance research in reinforcement learning, synthetic data generation, and agentic AI.
• Own and lead end-to-end research initiatives and guide teams while contributing technically.
• Develop rigorous benchmarks to measure reasoning, robustness, and generalization.

Connaissances

Python

PyTorch

JAX

Foundation models

Reinforcement learning

Research leadership

Outils

PyTorch

JAX

Description du poste

Company

Pramaana Labs is a frontier AI lab building the verification layer for AI. We turn complex human knowledge, including tax codes, legal rules, clinical guidelines and security protocols, into formal representations that support reasoning people can inspect and check. Our architecture pairs models trained to formalize and reason with a symbolic world model encoded in Lean 4. Founded in 2025 and headquartered in Palo Alto, Pramaana brings together researchers and formal methods experts from Google, DeepMind, and Glean.

The Role

We’re training models to formalize complex knowledge and reason with symbolic world models. As a Research Scientist focused on AI, you’ll develop approaches that improve the accuracy, coverage, efficiency, and generalization of our domain formalizer, autoformalizer, and prover.

Your scope spans model training, data quality, representations, and evaluation. You’ll work with an RL Scientist on reinforcement-learning approaches, Research Engineers on reliable experimentation, and Formal Methods experts on whether model outputs faithfully represent the source knowledge.

What You'll Do

• Lead ambitious research at the intersection of foundation models, formal reasoning, and verification, advancing Pramaana's mission to make AI reasoning reliable across complex, real-world domains.

• Develop novel model training and post-training approaches for translating complex domain knowledge into structured representations and enabling precise, verifiable reasoning.

• Drive hands-on research in reinforcement learning, synthetic data generation, reasoning, and agentic AI, exploring new techniques to advance model capabilities, generalization, and efficiency.

• Own and lead end-to-end research initiatives, identifying high-impact directions, setting technical strategy, and guiding teams of researchers and engineers while remaining deeply involved in the technical work.

• Develop rigorous benchmarks and evaluation methodologies to measure reasoning capabilities, robustness, and generalization, using experimental insights to drive fundamental model improvements.

• Partner closely with engineering to translate research breakthroughs into production capabilities, delivering measurable improvements in the performance and reliability of Pramaana's technology.

You Will Thrive in This Role If You Have

• A demonstrated track record of high-impact research, including research projects that have been successfully deployed and materially advanced a product, core technology, or organization's capabilities and direction.

• Deep hands-on experience in foundation model training and post-training, including one or more of large-scale pretraining, reinforcement learning, reasoning-focused training, or synthetic data generation.

• Strong technical depth in modern AI research, including frontier language models, agentic systems, multi-step reasoning, or structured prediction.

• Proficiency in Python and modern ML frameworks such as PyTorch or JAX, with experience implementing and scaling training experiments.

• Exceptional research judgment and experimental rigor, with the ability to identify promising directions, formulate hypotheses, design meaningful evaluations, and distinguish genuine advances from incremental gains.

• Demonstrated technical leadership and independence, with experience driving ambitious research initiatives, mentoring scientists, and collaborating effectively across research and engineering while remaining a strong individual contributor.

Especially Strong Candidates May Also Have

• Direct experience training or post-training frontier-scale models, including distributed training, large-scale reinforcement learning, or inference-time scaling.

• Research experience in neuro-symbolic AI, autoformalization, automated theorem proving, program synthesis, or formal verification.

• A strong research record demonstrated through influential publications, deployed research breakthroughs, or substantial improvements to state-of-the-art model capabilities.