Pramaana raises $27M to fix AI hallucinations with math

pramaana labs raised $27m in seed funding led by khosla ventures with participation from accel and boldcap to address ai reliability by integrating formal verification methods with large language models. the startup employs a dual-layer architecture where a conventional llm handles natural language processing while a deterministic verification layer, using languages like lean, mathematically proves the outputs correct. targeting verticals with codified rules such as tax law, cybersecurity, and drug discovery, pramaana is collaborating with domain experts including former irs commissioner danny werfel and professors from iit delhi and uc berkeley to formalize these complex systems. this approach attempts to solve the hallucination problem that currently prevents enterprises from deploying ai in high-stakes environments.