MUSE & DevJoy — AI Platform at Fannie Mae
Model lifecycle governance for institutional banking
Problem
Recoveries, Billing, and Liquidation workflows at Fannie Mae spanned disconnected legacy tools. Case and invoice submission took up to a week; liquidation and reconciliation ran on 8-week cycles. AI capabilities existed in the backend but had no consumable surface — model lifecycle governance, training, and TechOPS lived in engineering scripts, not in product.
Process
Research & journey mapping
Ran extensive user journey mapping and persona development with operators, model engineers, and ops teams. Facilitated cross-functional FigJam workshops to align business goals with user-centric AI solutions and identified the friction points that, once removed, drove a 62% reduction in user errors.
MUSE — multi-platform architecture
Architected MUSE as a unified model lifecycle governance surface across web, mobile, and connected device interfaces. Standardized complex backend workflows (model creation, modification, TechOPS) into one design system used by every internal AI team.
DevJoy — AI training UX
Designed DevJoy to let technical teams train banking models without leaving the product surface. Translated raw ML configuration into progressive, scalable interfaces — schema-aware, with safe defaults and audit-friendly diffs.
AI chatbot & document automation
Shipped a consumer-grade AI chatbot for institutional banking that approves documents with minimal human intervention. Iterative usability testing and empathy mapping validated high-fidelity prototypes for 100% alignment with user needs before engineering handoff.
Outcomes
Gallery
