Kepler builds a verifiable AI architecture for the financial services industry based on Claude.

In accurately extracting specific financial items from the 10-K financial reports of companies, Kepler's proprietary model has an accuracy rate of up to 94%; in comparison, relying solely on conventional industry-leading large models, the score is only between 38% and 46%. In response to the recent customer case study released by Anthropic, "How Kepler Leverages Claude to Build Verifiable AI for the Financial Services Industry," Kepler further disclosed its underlying technical architecture. The platform has currently deeply indexed 26 million Securities and Exchange Commission filings, 50 million other public documents, and 1 million private market documents, covering 14,000 companies in 27 major markets worldwide.