Real systems. Names withheld.

Problem

Film and TV data arrives from many sources, in many shapes, and rarely agrees. Teams needed one score they could trust to rank and compare titles.

What I built

Ingestion and normalization pipelines, ML scoring models, and the services that put those scores in front of the product.

Outcome

325,000 · bi-weekly · Improved content ranking and discovery

Problem

Keeping several channels supplied with on-brand posts, on schedule, and knowing what actually worked, takes a team's worth of hours every week.

What I built

Agentic workflows that draft posts, schedule them across channels and report on performance, with a person approving what goes out.

Outcome

10 accounts · 15 hours saved per week · 20% engagement increase

Problem

Municipal agendas, minutes and records are public, but they're buried in long documents that few people have time to read.

What I built

A system that collects agendas and records, pulls out the items that matter, and writes plain-language summaries of each meeting.

Outcome

52 municipalities covered · 10,200 meetings processed

Problem

Running client work and side projects means a constant stream of small tasks: planning, tracking and following up.

What I built

Agents built into my own management tooling that handle planning, tracking and follow-ups. I use them every day, which is the fastest way to find out what agents are actually good at.

Outcome

Handles planning, tracking and follow-ups · Saves several hours per week