AI Builder

A venture firm runs on judgment applied to too much information: thousands of companies, years of meeting records, signal arriving faster than anyone can read it. The partners didn't need a demo. They needed systems they could trust on a Tuesday morning, built by someone close enough to see how decisions actually get made.

I was embedded in the workflow, so feedback came directly from the partners using the systems. I could change something in days instead of waiting on an outside vendor's roadmap.

Four production systems, each built inside the workflow it serves:

1. A sourcing and scoring engine.Pulls from the firm's internal systems (CRM, email, meeting records, and similar) and scores opportunities against a taxonomy of 2,500+ companies. Surfaces what matters before anyone thinks to go look.
2. A tear-sheet engine for meeting prep.Walks into every partner meeting first. Compresses 4,000+ partner meeting notes into the two minutes of context that change how the conversation goes.
3. A traction monitor on a weekly cadence.Scans hundreds of companies every week for movement: hiring, shipping, raising. The partners stopped asking what's new and started asking what changed.
4. A verification layer.Catches hallucinated quotes and misattributed citations before anyone relies on them. Venture runs on trust, so the system that checks the system is the product.