Hello! I am Anthony and welcome to my page. I am currently working on the following:
An agent that turns long form videos and their transcripts into accurately timed clips, and re-clips only what actually changed.
- Model: Claude, via the Anthropic API, which decides which transcript segments need to be clipped.
- Tools: Whisperx for forced alignment, pyannote-audio for audio/video segmentation, ffmpeg for cutting, sentence-transformer embeddings for change detection.
- Loop: Each re-run diffs the transcript and decides whether to perform a patch or a full realign per segment.
- Backend: A modular Python package covering segmentation, curation, alignment, decisions, and caching.
- Interface: A FastAPI
/curateand/approvelayer for human review before anything is cut.
Automating a global mask catalog. One culturally researched mask a day, posted to X.
- Data: Already researched masks persisted in a database (PostgreSQL).
- Automation: A script reads the next unposted mask entry/row, attaches its image, and posts via the X API.
Contact me to connect on any project you might find interesting. Explore, fork, branch and share! Your input is welcome.

