Case studies
AI systems I have built.
Six builds: custom software a Northeast Florida contractor owns and runs its operation on, a live client investor portal, two shipped products of my own, and research-grade prototypes for high-stakes evaluation and reasoning. Real code in every one, no slideware.
Construction
Northeast Florida Commercial Contractor
Custom apps the contractor owns, replacing the rented tools its team used to fight. The system of record stays; the daily work moves into software they own.
An embedded engineer in the contractor's own cloud, building apps the company owns app by app: cost-to-complete forecasting, change-order routing, approvals.
Owned apps replacing rented per-seat tools
Read the case study →Real estate / fintech
Commercial Real-Estate Developer
One owned portal replacing a stack of SaaS subscriptions: investor management, payments, e-signature, and analytics for a capital raise.
A custom investor and stakeholder portal for a commercial real-estate developer raising capital, built to replace fragmented SaaS with one system they own, running on their own Google Cloud.
One owned portal replacing fragmented SaaS
Read the case study →Legal
TextTimeline
Retrieval-first evidence search for family-law text exports. Every result cites the source message.
A search engine for SMS evidence. An attorney asks a question in plain English; every finding traces back to a specific timestamped message.
100% citation coverage
Read the case study →Fintech
GammaRips
Autonomous overnight options-flow scanner. Picks one V5.3 contract per day with pre-set stop and target. Every paper trade published, win or loss.
Fourteen Cloud Run services, ~20 schedulers, and a multi-agent publishing layer that filters market noise into one tracked, public, mechanically-held trade idea per day.
One pick per day, fully published
Read the case study →Computer vision
Detection Evaluation Shell
Production evaluation infrastructure for an object-detection model: per-class metrics, calibration, drift testing, and latency, on an X-ray contraband benchmark.
An open-source evaluation shell I built around a public X-ray contraband detector: the part that tells you whether a detection model is actually trustworthy, not just what its headline score is.
Open-source eval shell, public on GitHub
Read the case study →Federal research
Compositional Verification Engine
A typed bridge between machine-learning outputs and symbolic reasoning, with bounded inference, explicit abstention, and provenance on every decision.
A Rust prototype I built for a federal research solicitation: it translates typed ML outputs into logical facts, runs bounded reasoning through a real symbolic backend, and abstains when a case falls outside what it can support.
10/10 integration tests passing
Read the case study →Writing
Why I build it this way.
The same patterns, written up: how the ownership math works and how I build retrieval that cites every source.
The ML was the easy part
A weekend pipeline that isolates a vocal, swaps the singer with a Retrieval-based Voice Conversion model, and renders a promo video, fully automated. The hard part was not the model. It was making a fragile, half-abandoned dependency stack run reliably across a Windows box and a Colab GPU.
Read the post →A content pipeline that can't go off-brand
I do not trust an AI to remember the legal disclaimer, so I do not ask it to. Deterministic code enforces the non-negotiables before any reviewer reads the draft. That is what makes unattended publishing safe to leave alone.
Read the post →Want similar work in your business?
Thirty minutes, no scripts. Describe the problem and I'll tell you if I can ship it.