Kepler (formerly Keru.ai)

Kepler, formerly known as Keru.ai, builds the verifiable AI infrastructure layer for industries where being wrong is not an option. Its approach eliminates AI hallucinations by separating language reasoning from deterministic computation: an AI layer interprets questions and constructs execution plans, while a deterministic code layer handles data extraction, calculation and consistency enforcement. The result is an answer that traces back to its exact source document, page and line item, with verifiable citations and reproducible output - asking the same question always yields the same answer.

Among its products, the Kepler Verifiable AI Platform implements this two-layer architecture across data retrieval and computation, with an emphasis on source traceability and reproducibility. The company also deploys a Financial Services Query System, production software that allows analysts in investment banking and financial analysis to query SEC filings, earnings transcripts and 10-Ks with full transparency.

Kepler's founding team brings more than 15 years of experience building mission-critical data systems at Palantir, Citadel and Meta. The engineering culture reflects that background: accuracy, verifiability and consistency are treated as non-negotiable, and transparency is enforced down to the document, page and line item. Financial services is the current focus, though the technology is aimed at any sector in which errors carry severe consequences.