Artificial intelligence is rapidly becoming indispensable for businesses, but reliability remains one of the biggest challenges. Errors, known as hallucinations, can cause immense harm, especially in critical sectors such as law, drug discovery, and tax preparation. To address this, a startup called Pramaana Labs has announced a $27 million seed funding round led by Khosla Ventures, with participation from Accel, BoldCap, Nexus Venture Partners, Premji Invest, and Unbound. The goal is to combine the power of large language models (LLMs) with the mathematical precision of formal verification.
Formal verification is a technique used for decades in software engineering to mathematically prove that a system behaves as intended. Pramaana Labs aims to apply this approach to AI, using the open-source programming language LEAN, known for verifying mathematical proofs. CEO and co-founder Ranjan Rajagopalan explained that while a traditional LLM offers flexibility in answering natural language questions, a deterministic layer on top ensures the responses are correct, especially in contexts where errors are unacceptable.
Sponsored Protocol
Applications in Critical Sectors
Pramaana will focus on highly sensitive verticals. In the legal field, an error in interpretation could have severe consequences. In drug discovery, hallucinations could lead to wrong conclusions about potentially lethal molecules. In tax preparation, a miscalculation could cost taxpayers dearly. For tax law, the startup is working with Danny Werfel, former IRS commissioner, while professors from IIT Delhi, IIT Madras, and UC Berkeley oversee the cybersecurity and drug discovery systems.
Sponsored Protocol
Rajagopalan emphasized that many complex problems are not unsolvable but simply unformalized. Once the rules are codified, reasoning becomes deterministic. This echoes projects like France's CATALA, which formalized much of the country's tax and benefit system into executable code. Pramaana plans to replicate this on a larger scale by building custom formal verification systems for each domain, overseen by subject matter experts.
The investment from Khosla Ventures and others comes as enterprises struggle to turn AI pilot programs into reliable operational features. The ability to combine flexibility and rigor could be a game-changer. Other players are exploring similar paths, such as Stanford's DeLM, which cuts multi-agent coordination costs without a central orchestrator, and Sakana Marlin, an AI agent that writes complex strategic reports in hours. For more on these innovations, check the articles on Sakana Marlin and Stanford's DeLM.
Sponsored Protocol
Formal verification is a well-established branch of computer science, as described on Wikipedia. Pramaana Labs is adapting it to AI, paving the way for a future where intelligent systems can be considered as reliable as mathematical calculations. With $27 million and a team of experts, the startup aims to turn this vision into reality, one domain at a time.
