Underwriting Automation Platform
ShippedA system that bridges our Excel-based underwriting models with the data warehouse, automating data entry and setting the foundation for AI-assisted underwriting.
The Problem
Our underwriting process required analysts to manually transfer data between sophisticated Excel models and multiple downstream systems. The models were brilliant but isolated. Every deal required hours of manual data entry. The goal was to automate the bridge so models could run on-demand and results could flow into the data warehouse automatically.
For weeks, this felt like a fantasy. I was stuck in a two-front war.
Backend: An Office Script engine running through Power Automate. Slow, fragile, timing out constantly on circular-reference-heavy iterative calculations.
Frontend: A brittle dirty-tracking system watching 500+ fields for changes. Every bug fix spawned two more. Progress stalled.
The breakthrough came when I stopped patching cracks. I partnered with AI — Gemini as cognitive partner for architecture, Claude Code as coding partner for implementation. We stress-tested alternatives, validated architectures, rebuilt the foundation.
A new engine. A .NET microservice using Aspose.Cells, deployed on Google Cloud Run. Built to handle iterative calculations at scale. Before: a 2-minute timeout. After: 9.85 seconds.
A new brain. A frontend rebuilt on Zustand and Immer. The fragile UI became predictable and stable.
The wall is gone. The bridge to the data warehouse is real. Automation is unlocked and the stage is set for AI in the underwriting workflow itself. Building is hard. Persisting is harder.
Essays about this build
AI has two modes: Cognitive and Operational. Work slop comes from treating everything like Cognitive Mode. The fix is knowing when to constrain AI to data-backed execution.
For weeks, the dream of bridging our underwriting models to the data warehouse felt like a fantasy. The breakthrough came when I stopped patching cracks and rebuilt the foundation.