At Meridial, I compare LLM outputs for factuality, accuracy, and relevance, applying annotation guidelines to create preference data for RLHF. I also identify hallucinations, inconsistencies, and bias, and write structured justifications for engineering teams.
At Turing, I completed 100+ approved submissions, including French audio diarization, transcription correction, and paralinguistic annotation on the NVIDIA Eagle project. I was selected as an Audio/Video Annotation Trainer (French Language), supporting guideline calibration and peer quality reviews.
