**Summary:** On May 21, 2026, Kepler, a verifiable AI platform for financial research, expanded on an architecture detailed in an Anthropic customer profile titled “How Kepler built verifiable AI for financial services with Claude.” The profile describes a system combining language models with traditional financial software to meet the rigor financial workflows require, making answers defensible in an IC memo or under audit. Kepler’s architecture splits work: language models interpret questions and generate narratives, while code retrieves data, runs calculations, and renders citations, mediated by a financial ontology mapping terms like “EBITDA” to exact line items. Kepler’s specialized models read a 10-K and identify the right line item 94 percent of the time, compared to frontier models scoring 38 to 46 percent. The platform indexes 26 million SEC filings, 50 million additional public documents, and 1 million private documents across 14,000 companies and 27 global markets. Dr. John McRaven, CTO of Kepler, stated the model is one stage in a pipeline, while CEO Vinoo Ganesh noted Claude consistently held the plan together. For buy-side analysts, every figure clicks through to its source, calculations are reproducible, and outputs are auditable end-to-end.