Al Data Annotator (Part-Time Contract)
Alignerr (LabelBox)
Mar 2025 - Current (1 year 6 months)
Acted as an attempter in a few Generalist tasks. Evaluated Al-generated outputs in native Tamil script, transliterated formats, and English to ensure linguistic precision, cultural relevance, and technical accuracy across STEM-related content. Worked as Voice Activity Detection (VAD) Data Annotator. Accurately transcribed and labeled user and assistant speech using a specialized set of token types. This work was critical in training and evaluating speech-recognition (ASR) models, specifically improving their handling of nuanced speech elements like filled pauses (e.g., "ums"), barge-ins, and acknowledgments.