At Welodata, I review evaluations of AI-generated responses for linguistic correctness, naturalness, formatting, and accuracy. I check that scores and justifications match the issues identified and request revisions when needed.
At Innodata India Pvt. Ltd., I evaluated and annotated AI-generated text, image, and video content using structured guidelines. I assessed outputs for factual correctness, relevance, coherence, and policy compliance.
At CrunchLabs, I translated and localized 200+ technical and educational content pieces from English to Urdu, preserving clarity, cultural relevance, and technical intent. I also used AI-assisted tools for dubbing, timing alignment, and multilingual adaptation.
For my Aahang Search project, I developed a context-aware retrieval system for Urdu and Indic poetry that handles fragmented verses, spelling variations, typos, and transliteration noise. I implemented TF-IDF, cosine similarity, character-level n-grams, and SQLite-based storage.
