At micro1, I evaluate frontier multimodal model outputs against project guidelines, decision trees, and data-integrity rules. I flag hallucinations, timestamp errors, and safety-boundary violations with reproducible written feedback.
At Innodata Inc., I annotated short-form video, audio, and images for Meta content-understanding programs, verifying AI-generated descriptions against Source of Truth guidelines. I scored 35/35 on the Content Understanding Video Master evaluation and was promoted to QA Super-Rater, where I audited peer submissions and resolved guideline edge cases.
With AfterQuery, Handshake AI, and Outlier, I rated generative image and video outputs across rubric dimensions and audited multimodal image records for grounding and unsupported claims.
As a freelance developer, I built and run FlickCollection, which reached 145,000+ monthly pageviews within nine months of launch. I also built SoundsSwapp, a Spotify-to-YouTube playlist converter that matches tracks at 98%+ accuracy across 5,000+ tracks.
