How Kepler Built Verifiable AI for Financial Services — Vinoo Ganesh
Kepler built verifiable AI for financial services where evals alone cannot guarantee trustworthy outputs
“You cannot take a non-deterministic LLM and eval your way to something deterministic. These are probability machines.”
Kepler CEO Venu Ganesh argues the financial services AI conversation is too focused on output volume and not enough on verifiability, contending that evals cannot make non-deterministic LLMs trustworthy for high-stakes work. Kepler, the only company Anthropic has published a case study on, built a system for verifiable AI outputs in finance. The core insight—that AI democratizing content production destroys informational edge (alpha decay)—has broad implications for how financial firms should think about AI differentiation.