Join a workshop
← Back to all work
Hero image: terminal screenshot or architecture diagram
AI & Automation Active

Luma Data Pipeline

Automated event analytics across 11 events and 2,100+ guests, replacing manual spreadsheet tracking with a CLI-driven SQLite pipeline.

Role Solo builder
Timeline May 2026 — Present
Stack Python, SQLite, Claude API, Click

Context

After running 10+ AI Build Club events, I was tracking attendance, feedback, and guest data across disconnected spreadsheets and Luma exports. Every time I needed a metric for a partnership pitch or grant application, I spent hours pulling numbers manually. The data existed but was scattered and stale the moment I exported it.

My Role

I designed the schema, built the CLI, wrote the import pipeline, and maintain it. This is a solo project I use daily to run AI Build Club operations.

Approach

I needed something that fit my workflow: a CLI tool I could run from the terminal, not a dashboard I'd forget to check. Key decisions:

  • SQLite over Postgres — single-file database, no server to maintain, portable across machines.
  • CLI over web UI — I live in the terminal. A web dashboard would have been an extra thing to maintain.
  • Luma API as single source of truth — pull fresh data on demand rather than maintaining manual imports.
Architecture diagram: Luma API → Python CLI → SQLite → Reports

How It Works

The pipeline pulls event and guest data from the Luma API, normalizes it into a relational schema (events, guests, feedback), and stores everything in SQLite. CLI commands expose queries for common operations: attendance by event, guest overlap, feedback aggregation, and export for reports.

terminal 
$ luma sync --all
Syncing 11 events...
  ✓ 2,147 guests imported
  ✓ 847 unique attendees
  ✓ 156 feedback responses

$ luma stats --event "march-2026"
  Approved: 89  |  Showed: 52  |  Show rate: 58%
  New guests: 31  |  Returning: 21

Results

2,100+ Guest records processed
11 Events tracked
~4hrs Saved per reporting cycle

The pipeline replaced a manual process that took half a day of spreadsheet work every time I needed numbers for a partnership pitch, grant application, or event retrospective. Now it takes one command.

What I Learned

Placeholder for reflections. What surprised you, what you'd do differently, what this project taught you about building tools for yourself vs. for others.