At Atlas Capture, I annotated 5,000+ short videos of human activity against written directives, achieving an 80% quality score. I used detailed rubrics to assess action boundaries, completeness, and description quality.
I also refined AI-generated annotations by comparing video samples against instructions, and documented decisions and edge cases to support reviewer consistency.
On my independent Docta ODI project, I built and maintain an autonomous ML prediction API, migrating its data flow from Excel to a WAL-mode SQLite database. My work across healthcare and visa pathways has also involved precise, standardized documentation in regulated and complex settings.
