Founded by Ranjan Rajagopalan, Krishnan Raghavan, and Sanjay Ganapathy Subramaniam, the startup converts specialized knowledge from areas such as tax codes, clinical guidance, legal regulations, and safety frameworks into representations that can be validated through machine-based verification.
The company’s approach draws on formal verification, a methodology widely used in fields such as semiconductor design and aerospace engineering to ensure systems behave according to predefined rules. Pramaana is applying similar principles to AI reasoning, with the goal of making outputs traceable and verifiable rather than relying entirely on model-generated responses.
According to the company, its system combines conventional large language models with a deterministic verification layer. The AI model handles natural language inputs and complex reasoning tasks, while the verification component evaluates whether the output aligns with codified rules.
“It’s like math in the sense that you have a lot of rules that you need to abide by,” Rajagopalan said when describing tax regulations. “Once you have a codified version of it, the reasoning on top of it starts becoming deterministic.”
Pramaana’s verification framework is based on formal methods that use the open-source LEAN programming language, a system commonly employed to verify mathematical proofs. For each industry it serves, the company develops a dedicated verification system with oversight from subject-matter experts.
The startup is already working with former IRS Commissioner Danny Werfel on tax-related applications, while academics from IIT Delhi, IIT Madras, and UC Berkeley are involved in areas including cybersecurity and drug discovery.
Pramaana recently hosted its first Verification Summit in San Francisco, an event that featured Khosla Ventures founder Vinod Khosla.
The funding reflects growing interest in technologies designed to improve AI reliability as organizations move beyond experimentation and begin deploying AI systems in operational environments. Pramaana is positioning itself as infrastructure rather than a model developer, focusing on verification tools that sit alongside AI systems rather than competing with them.
“The world’s hardest problems are not unsolvable. They are unformalized,” Rajagopalan said. “Every domain where being wrong can cost someone their health, money, or freedom has rules.”
By translating those rules into machine-verifiable systems, Pramaana aims to create AI workflows that can demonstrate how conclusions were reached and whether those conclusions satisfy the requirements of the domains in which they operate.
This analysis is based on reporting from Crypto Briefing.
Image courtesy of Pramaana Labs.
This article was generated with AI assistance and reviewed for accuracy and quality.
