Pramaana Labs raises $27M to bring formal verification to AI

Read the original article →

What happened

Pramaana Labs announced $27 million in seed funding on June 17, led by Khosla Ventures, with Accel, Nexus and Premji Invest joining. Led by co-founder and CEO Ranjan Rajagopalan, the company aims to bring formal verification to AI in high-stakes verticals such as law, drug discovery and tax preparation.

Its system pairs a conventional LLM with a deterministic formal-verification layer built on the LEAN proof language, constructing a verification layer per vertical to reduce hallucinations. The company cites France's CATALA project, which encoded tax law as verifiable code, as precedent for applying formal methods to domains that demand precise, checkable answers.

Why it matters

Formal verification offers mathematical guarantees that ordinary LLM outputs cannot provide, which matters most in law, medicine and tax where errors are costly. Using an established proof language like LEAN grounds the approach in decades of formal-methods research.

It reflects a broader shift toward pairing probabilistic models with deterministic checks rather than trusting models alone.

MintedBrain take

Formal verification is powerful but labor-intensive, since each vertical needs its rules encoded before the layer can check anything. The guarantees are strong only within what has been formalized, so probe where coverage ends before relying on the system for consequential decisions.

References

This article was originally published at The Next Web. For the full piece, read the original article.

Discussion

  • Loading…

← Back to News