EST. 2026 · MILWAUKEE, WI · 47.2°N
Production AI for Midwest architecture and real estate firms.
Custom systems, shipped in weeks, not quarters. Built once, maintained, and yours.
§ 01 · The problem
We build the part that ships.
95%
of enterprise GenAI pilots produce no measurable contribution to revenue or margin.
MIT NANDA · State of AI in Business · 2025
80.3%
of enterprise AI projects fail to deliver business value. Twice the failure rate of conventional software.
RAND Corporation · confirmed by Gartner · Apr 2026
39%
of companies now have AI in production at scale, up from under 5% two years ago. The gap is execution, not models.
MIT Sloan Management Review · 2026
§ 02 · Selected work
Brokerage · Tenant intelligence
A tenant-intelligence platform for one brokerage team, covering industrial activity reporting and tenant intelligence, that replaced their prior system entirely. The follow-on scope: template automation and automated broker PowerPoint generation.
Outcome
Replaced their existing tooling, then brought me back to build the next thing.
Pharmacy · Finance & fraud
A financial aggregation layer pulling 20+ bank accounts and 20+ credit cards across Wisconsin and the Midwest into one place, with an AI analyst on top so anyone can ask a question across any account, bank, or card without pulling statements by hand.
Outcome
Siloed finances unified, plus fraud and anomaly detection across wholesaler volume and drug data.
Architecture · 100-person firm
A retrieval system over every municipal code the firm regularly works in, with an AI layer that finds the relevant sections and cites the exact code it drew from.
Outcome
Multi-hour zoning lookups cut to about 27 minutes. In active use, unchanged since handoff.
Read the full case study ↓§ 03 · Case study · 2025
Problem ·A 100-person Milwaukee architecture firm works across most of the municipalities in southeastern Wisconsin. Every early-stage project starts with the same task: pulling zoning, setback, and overlay-district rules out of a different city’s code. Done by hand, that’s a multi-hour stretch of careful reading before anyone can sketch.
Build ·I shipped a retrieval system over every municipal code the firm regularly works in, with an AI layer that finds the relevant sections and summarizes what they actually require. Every answer cites the exact code section it drew from, so the architect can verify in one click.
Outcome ·Multi-hour lookups now take around twenty to thirty minutes. The tool is in active use across the firm and has held up unchanged since handoff.
“I was in a rush the other day and used Win’s tool. What would have taken me an hour and a half took twenty-seven minutes. I timed it.”
§ 04 · Who builds this
An AI engineer at Associated Bank, he joins Baird as an investment-banking AI engineer in August 2026. He builds production AI systems for Midwest architecture, engineering, and real estate firms, and serves as president of AI Forge while studying in Marquette’s AIM Fintech program. His research on a next-generation telemetry-monitoring system for astronaut safety is being published and presented at ICES 2026 in Puerto Rico.
Lapham Labs is a one-person, full-stack practice. I do the discovery, the design, the engineering, and the hand-off. When a build needs more capacity, I bring in vetted engineers from AI Forge, the AI engineering organization I lead at Marquette, and call on my advisor Hunter Sandidge for senior expertise. Every line of code still ships through me.
Research
A next-generation telemetry-monitoring system for astronaut safety.
Published · Presented at ICES 2026 · Puerto Rico
Read the paper (PDF) ↗Advisor
Hunter Sandidge
Former Manager of Data Science & AI Engineering, Collins Aerospace
§ 05 · FAQ
Thirty minutes. No deck. No pitch. Just a real conversation about what you’d ship if you had the team.