Radio fan.
Self-taught dev.
Same brain.
I spent years as a radio listener wanting to go back and listen to older episodes of my favorite morning show. None were saved. When the show got cancelled, they were gone for good — and I didn’t have a single clip. When a new edition started up, I made sure that wouldn’t happen again. I started saving every MP3.
But saving files isn’t the same as being able to find anything in them. What did they talk about on March 8th? When did they mention the Cowboys draft? I had hundreds of hours of audio and no way to search any of it. So I built one. What started as a personal archive turned into a full pipeline — transcription, speaker diarization, vector embeddings, semantic search — running automatically every weekday.
That’s the pattern. I find a coordination problem that shouldn’t exist, and I build the system to fix it. The podcast pipeline does it for audio. C.R.E.W. does it for emergency robotics. Both are open source, both are running, and both started from a simple frustration that nobody else had solved.