proprise (yc s23) · 2024–2026 · founding product & customer success

primer: ai agents that underwrite deals

Grew to 60% of company revenue, driving ARR from ~$150K to over $700K. Used by 50+ firms managing $150B+ in commercial real estate.

Every underwriting interview surfaced the same complaint: analysts were retyping numbers from offering memos, rent rolls, and P&Ls into Excel, deal after deal.

While our CTO was traveling, I built the first version of Primer myself, an Excel plugin that could underwrite real deals. We had a paying customer within a month.

From there I ran product and stayed every customer's primary technical contact. I designed in Figma, spec'd features in Linear, and shipped thousands of lines of production code with coding agents.

8–10 extraction agents per customer, built for their own document formats
Every agent tested against real deals before the customer went live; 99%+ extraction accuracy
Ten eval passes per template, since most errors were variance, not repeated mistakes
Cheap models for simple extractions, frontier reasoning for messy files: cost and processing time both halved
Review screens put the model's reasoning beside each answer, with one-click corrections it remembered
Customers tripled underwriting volume; one cut per-deal work from 90 minutes to 10