By the time a lead waits 10 minutes for a callback, you're 10× less likely to ever reach them. After an hour, 60× less likely.
I build AI systems that respond to your leads, reviews, and calls instantly, automatically, even while you sleep. No pitch, just a look at your current setup and where AI can cut your response time and get you clients.
A lead fills your form. Within 60 seconds, they get a personalized AI text. You get the notification. No agent needed, no lead lost.
Every Google review — 1-star or 5-star — gets a thoughtful, on-brand draft response in your Slack or email within minutes. You approve, copy, paste.
A trained AI answers your phones after hours, qualifies callers, answers FAQs, and routes urgent calls — so you never miss a client because you were in a showing.
Build software to last.
Deliberate, elegant, exposed.
A web app for comparing apartment listings side-by-side. Users paste the full text of any listing and the LLM-backed parser pulls structured data like rent, beds, baths, amenities, and available lease terms and constructs the data as compact cards in a single comparison view with interactive map and personalization filters, regardless of which site the listing came from.
The instinct with AI implementation is to collect everything and show everything. The challenge is deciding what's actually worth showing.
A competitive league platform with brackets, ratings, live standings, and accounts. Working as lead design engineer to translate the product vision directly into HTML, CSS, and JavaScript, with an emphasis on keeping the dense statistical depth accessible without overwhelming new users. Currently in public beta testing.
Designing inside someone else's vision is mostly an exercise in earning the right to say "less." The work is in knowing when you've earned it.
An interconnected home automation and AI system. Designed in Oberon as a deliberate constraint to test how much sophistication a minimal foundation can carry and as a study in organic control over localized hardware systems. The architecture treats devices as participants in a shared conversation rather than as instruments, closer to an ecosystem than a control system.
Cooperative systems are just systems whose hierarchy has been earned.
A simulation substrate for evolutionary computation in Rust, populating a 'terrarium' environment with colonies of communicating agents. The architecture treats the colony, and not the individual as the unit of selection. A Queen evolvable autoencoder compresses the problem space, and specialized drones solve the compressed representation, with strategies emerging from their interaction rather than from any central design. The base terrarium and genome system are working; the colony hierarchy and adaptive compression are the active research frontier. A design paper is in progress.
Single answers hide trade-offs; real problems live on a Pareto frontier.
Book a free 20-minute call or contact me directly.