With SOL Lead Engine, I built a daily lead pipeline for appointment businesses around Austin, Texas. It finds local businesses through OpenStreetMap, deduplicates them in PostgreSQL, and audits their websites from a customer’s perspective.
I wrote deterministic Python scoring rules, with every point explained in strict rubrics and covered by 11 unit tests. Leads are sent to Telegram for human review, decisions are recorded in an append-only log, and failed runs trigger alerts with their cause.
For Business Brain, I built a retrieval-augmented knowledge base over a 90-note Obsidian vault. It re-embeds edited files into PostgreSQL with pgvector, answers only when retrieved notes meet a similarity threshold, and cites its sources.
I also improved a stalled run from 49 minutes to 38 seconds and added weekly self-tests and daily health checks. My operations background shapes how I approach automation: deterministic rules, human-in-the-loop approval, error handling, and clear documentation.
