Pramaana Labs raises $27M to make AI prove its answers

Pramaana Labs Inc., a formal verification startup, announced a $27 million seed funding round to advance its AI verification system. The platform acts as a compiler for high-stakes AI applications, cross-checking answers against domain-specific rules before approval. Targeted industries include tax compliance, medical diagnostics, cybersecurity, and financial regulations, where errors carry significant costs. The company leverages Lean, an open-source language for machine-checked proofs, to formalize rules from fields like the U.S. tax code or clinical protocols. When a user submits a query, the system restates it as a formal claim and verifies it through a proof engine. If the answer fails validation, the system highlights the violated rule, ensuring accuracy in regulated workflows. Pramaana’s founders—three alumni of the Indian Institute of Technology Madras—bring expertise from Google, DeepMind, Meta, and UC Berkeley. The team includes researchers from top tech firms, while domain experts like former IRS Commissioner Danny Werfel advise on tax applications. Professors from IIT Delhi, IIT Madras, and UC Berkeley oversee cybersecurity and drug discovery efforts. The funding, led by Khosla Ventures with participation from Accel, BoldCap, and others, will support model training, hiring, and expanding domain-specific verification systems. The company cites Google DeepMind’s AlphaProof and France’s Catala project as precedents for automating formal verification in regulated sectors. Pramaana’s approach aims to bridge the gap between AI-generated answers and deterministic proof, addressing a critical need in fields where compliance and precision are non-negotiable.