Labs
Smaller builds that show the range: a serverless AI agent and a domain-specific quoting tool.
The problem
Two different itches: recruiters want a fast, honest read on fit before a call; and contractors waste real money on inaccurate materials takeoffs. Both are solvable with a small, sharp tool.
What I built
Jackson-bot is a serverless AI agent (Cloudflare Workers + Claude Sonnet 4.6) that answers recruiter questions about my background 24/7, with conversation memory, KV-backed rate limiting, anonymized logging, and role-aware resume routing. Waterfront is a React + Vite quoting tool that turns a decking/seawall job spec into an accurate materials takeoff with waste math and a branded customer quote.
How AI was leveraged
Jackson-bot is AI-native — the entire business logic lives in a 177-line system prompt and Claude does all the conversational reasoning while the Worker handles safety, memory, and routing.
Engineering highlights
- Jackson-bot: 30 req/IP/hr rate limiting in Cloudflare KV, IP-anonymized logging, 4 role-specific resume variants behind a /resume redirect
- Jackson-bot: 14-message conversation memory with a voice-tuned system prompt
- Waterfront: domain formulas (joists, decking rows, fasteners, leftover math) isolated in a rules module
- Waterfront: localStorage-persisted pricing config and print-to-PDF / CSV customer quotes
Inside the venture
A Claude-powered agent on Cloudflare Workers that represents me to recruiters 24/7.
A React quoting tool that produces accurate materials takeoffs and branded quotes for a decking/seawall contractor.
Outcome
Both shipped. Together they show serverless AI engineering and pragmatic domain tooling.
Road Ready Safety