Make messy inputs usable.
I build systems end to end. Not specs. Not decks. Actual working infrastructure: pipelines that route leads, automations that replace 25 hours of manual work per week, attribution stacks that show where revenue actually comes from.
The hacker in me gets bored designing things I'm not also building. Thinking and doing happen in the same motion or they don't happen at all.
Most data projects fail before the analysis stage. The raw input is messy, inconsistent, or missing context. The first 80% of the work is cleaning, tagging, and deciding what a row actually means.
I built conversation mining pipelines that take unstructured text, anonymize it, extract themes, and output usable drafts. Not magic. Just structured thinking applied to a messy problem.
The best automation is the kind nobody notices. Lead hits the CRM, gets scored, routed, followed up. No one clicked anything. The system just knew what to do. Building that invisibility takes longer than building the visible parts.
Python · JavaScript · Astro · FastAPI · React · Node.js
PostgreSQL · Redis · Supabase · Vercel
HubSpot · Zapier · Google Analytics · Meta Ads API
"The best way to predict the future is to invent it." — Alan Kay