Portfolio managers at Waterfall Asset Management evaluate hundreds of loans daily and track performance on Databricks. But loan “tapes” arrive as spreadsheets in high volume and inconsistent formats across originating institutions, making bids dependent on time-consuming data cleaning and harmonization before standard metrics and dashboards can be computed. Waterfall Asset Management and Prophecy built a specialized Claude Code “harmonization” agent that automates onboarding into a common schema, requiring only a final validation pass where users review AI confidence, reasoning, and lineage and make any edits. In production, WFAM achieved 93% first-pass accuracy—now approaching 100%—reducing reliance on data experts and materially accelerating investment decisions.
Leave a comment