← back

How Kepler Built Verifiable AI for Financial Services — Vinoo Ganesh

10.2K views · Jul 29, 2026 · 22:30 min · Watch on YouTube ↗
Takeaway

Financial AI work products need traceable sources and deterministic validation of numerical transformations.

Summary

  • Kepler frames AI adoption in finance as a verification bottleneck: generating analyses is easier than trusting their numerical content.
  • A citation identifies a source but does not prove that a value was extracted or transformed correctly.
  • Verification must reflect an organization’s trusted sources and permitted transformations, even when firms reach different investment conclusions.
  • The proposed architecture augments probabilistic reasoning with deterministic computation and introduces atomic provenance as a platform principle.
keplerfinancial-aiprovenance
Original description
In finance a number is worthless until you can say where it came from. Vinoo Ganesh, CEO of Kepler, starts from the fact that language models are probability machines, brilliant at next token prediction and unreliable at the deterministic work, like arithmetic, that finance actually runs on. So Kepler treats the model as one part of a system rather than the whole answer, wrapping it in a deterministic substrate that makes every figure traceable.

It rests on three tenets. Atomic provenance means every number is tied to its source and stripped out if it cannot be independently verified. Scope determinism keeps the model on the nondeterministic tasks it is good at and pulls the actual figure, say revenue from a 10K filing, deterministically behind the scenes, with reconciliation on top. The third is treating every extracted number like a pull request that gets reviewed, so entities are caught and nothing is invented. The result is a grounded system where the edge comes not from producing content but from verifying it.

Speaker info:
https://x.com/vinooganesh
  / vinoo-ganesh  
https://vinoo.io

Timestamps:
0:00 - Introduction: a data background in finance
1:42 - Why trust and verifiability matter now
2:57 - Models are probability machines
4:27 - Why analysts still put in the hours
8:22 - Modeling AI like an overworked VP
9:26 - Atomic provenance
12:01 - Scope determinism
13:52 - Reconciliation and pulling real numbers
15:06 - Extracting entities without misses
16:34 - Toward zero invented securities
20:31 - Where a number really comes from