As companies struggle to turn AI pilot programs into functional parts of their businesses, trust is taking center stage. A new startup hopes to solve this problem by leveraging the tools of mathematical formalization, combining one of computer science’s most reliable and one of its most chaotic systems.
On Wednesday, Pramaana Labs announced $27 million in seed funding led by Khosla Ventures with participation from Accel, BoldCap, Nexus Venture Partners, Premji Invest, and Unbound.
Pramaana will focus on highly sensitive areas such as law, drug discovery, and tax preparation. In these fields, errors are costly and reliability is at a premium. Introducing AI into these systems will require greater protection against illusions and errors than currently exists. But they also have a unique aptitude for formalization, according to Ranjan Rajagopalan, co-founder and CEO of Pramaana.
“It’s like math in the sense that there are a lot of rules that need to be followed,” Rajagopalan told TechCrunch about the rules of the tax code. “Once you create a codified version, the reasoning on top of it starts to become deterministic.”
Pramaana’s system still runs on traditional LLM, giving it the flexibility to answer natural language questions and tackle complex problems that traditional computers can’t handle. But on top of that LLM there is a deterministic layer that ensures that the work of the LLM checks out.
This combination of LLM engine and deterministic verification is a common setup. Pramaana’s unique approach is to use formal verification tools powered by the open-source LEAN programming language, which is used to verify mathematical proofs. Much of this research actually has precedent. Mr. Rajagopalan points to France’s CATALA project, which formalizes much of the country’s tax and benefit system into executable code.
For each use case, Pramaana builds a unique LEAN-style formal validation system overseen by domain experts. On tax law, the company is working with former Internal Revenue Service Commissioner Danny Wuerffel, and professors from IIT Delhi, IIT Madras and the University of California, Berkeley, oversee cybersecurity and drug discovery systems.
“The world’s toughest problems are not unsolvable; they are unformalized,” says Rajagopalan. “There are always rules in areas where getting it wrong could cost someone their health, money, or freedom.”
All that remains is to codify these rules.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
