Kepler has created a groundbreaking platform that indexes over 26 million SEC filings, earnings call transcripts, and other financial documents, establishing a trust and verification layer for AI in the financial services sector.
In a heavily regulated industry, the need for auditable and accountable reporting is paramount. Financial firms require every figure in their reports to be verifiable against source documents. Traditional tools can extract data but often leave the verification process to analysts, which can lead to errors.
Vinoo Ganesh and John McRaven, with their extensive experience at Palantir, recognized this gap and sought to build a solution. After consulting with 147 financial firms, they found a common concern: while many wanted to leverage AI for research, trust in the output was lacking. This led to the development of Kepler, which integrates Claude as a reasoning layer, allowing analysts to pose questions in plain English and receive verifiable answers.
How Kepler Works:
- Kepler's platform can handle complex financial queries, ensuring accuracy through a structured approach.
- It employs a deterministic infrastructure that maintains data integrity throughout the analysis process.
- Claude is utilized for interpreting queries and managing ambiguity, enhancing the reliability of the outputs.
For instance, when an analyst requests specific inventory metrics over multiple quarters, the system identifies the necessary calculations and data sources, ensuring all fiscal periods are accurately represented.
Key Features of Kepler's Platform:
- Deterministic execution environments for critical calculations.
- A proprietary ontology mapping financial concepts to precise definitions.
- Customizable workflows for common financial tasks, ensuring consistent outputs.
The platform also incorporates rigorous testing protocols, evaluating every prompt and model upgrade against thousands of scenarios to ensure reliability before deployment.
Future Prospects:
