Curbing AI Errors: Pramaana Labs Raises $27 Million

As AI technologies evolve, large enterprises face reliability issues when implementing these systems into real business processes. Aiming to solve this problem, the startup Pramaana Labs has raised $27 million in an investment round led by Khosla Ventures. The round also included prestigious funds such as Accel, Boldcap, Nexus Venture Partners, Premji Invest, and Unbound. This was reported by Techcrunch.com news says.

The primary goal of Pramaana Labs is to eliminate "hallucinations" (fabricating incorrect information) and errors in systems by combining AI with mathematical precision. According to TechCrunch, the startup will direct the investment toward developing secure AI systems for sectors where errors are costly, such as law, pharmaceuticals, and tax reporting.

Mathematical Precision and Deterministic Approach

While currently popular LLM (large language models) are distinguished by their flexibility, they can often produce illogical results. Pramaana Labs is installing a special deterministic layer on top of these models. This layer is based on the open-source LEAN programming language used for verifying mathematical proofs, controlling AI responses based on strict rules.

Company founder and CEO Ranjan Rajagopalan emphasized that tax codes or legal norms have strict rules, much like mathematics. "If you have a codified version of these rules, then logical conclusions become precise and immutable," he says. This approach ensures that AI results are legally and scientifically correct while maintaining its ability to communicate freely.

Expert Oversight and Future Plans

Pramaana Labs is creating separate LEAN-style verification systems for each sector. The project has attracted some of the world's most prestigious experts. For example, Danny Werfel, former commissioner of the US Internal Revenue Service (IRS), serves as an advisor. Work in cybersecurity and pharmaceuticals is overseen by professors from IIT Delhi, IIT Madras, and UC Berkeley.

In the context of Uzbekistan, such verification systems are crucial when implementing AI into government services, tax, and banking systems. Because in financial and legal matters, even a small error made by AI can lead to serious consequences. Startups like Pramaana Labs aim to break exactly this wall of distrust.

According to Ranjan Rajagopalan, the world's most difficult problems are not unsolvable, but rather unsystematized. "In any field affecting human health, finances, or freedom, rules exist. Now our task is to turn these rules into code," he concludes. This technology will allow enterprises in the future to use AI not just as an experiment, but as a primary work tool.