Pramaana Labs Secures $27M Seed Funding from Khosla Ventures to Advance AI Formal Verification
As companies face challenges in integrating AI pilot programs into their operations, ensuring reliability has become a key focus. A new startup aims to tackle this issue by leveraging mathematical formalization, merging one of computer science’s most dependable systems with its more unpredictable aspects.
Pramaana Labs recently announced that it has secured $27 million in seed funding, spearheaded by Khosla Ventures, with contributions from Accel, Boldcap, Nexus Venture Partners, Premji Invest, and Unbound.
The startup intends to concentrate on critical sectors such as law, drug discovery, and tax preparation, where mistakes can lead to significant consequences and reliability is crucial. Implementing AI in these areas necessitates enhanced safeguards against inaccuracies and hallucinations that are currently lacking. However, co-founder and CEO Ranjan Rajagopalan believes these fields are particularly amenable to formalization.
“It resembles mathematics because there are numerous rules that must be followed,” Rajagopalan stated, referencing the complexities of tax law. “Once you have it properly codified, the reasoning applied becomes consistent.”
Pramaana’s solution relies on a standard large language model (LLM) that allows for natural language inquiries and addresses intricate issues that standard computers cannot manage. However, a deterministic layer is integrated above the LLM to ensure the accuracy of its outputs.
The combination of an LLM with a deterministic verification layer is a common methodology; what sets Pramaana apart is its application of formal verification techniques, utilizing the open-source LEAN programming language to validate mathematical proofs. Rajagopalan references France’s CATALA initiative as a successful example, wherein much of the country’s tax and benefits framework is converted into executable code.
For each application, Pramaana will develop a custom LEAN-like formal verification system, guided by subject matter experts. In the realm of tax law, the startup is collaborating with former IRS commissioner Danny Werfel, while faculty from IIT Delhi, IIT Madras, and UC Berkeley are involved in overseeing projects related to cybersecurity and drug discovery.
“The toughest challenges are not beyond solving; they are simply unformalized,” Rajagopalan asserts. “Every sector where errors can jeopardize a person’s health, finances, or freedom operates under a set of rules.”
Now, it is essential to codify those rules.
