Behavioral Data Annotation for LLM Training and Evaluation
Contributed to large-scale language model training and evaluation projects focused on behavioral health, conversational AI, and psychological reasoning tasks. Responsibilities included: Annotating and rating AI-generated responses for safety, clinical appropriateness, coherence, and factual accuracy. Classifying emotional tone and psychological constructs in text data (e.g., anxiety indicators, cognitive distortions, affective states). Reviewing and refining prompt-response pairs for supervised fine-tuning (SFT). Conducting structured text summarization and diagnostic-category tagging based on clinical descriptions. Evaluating responses involving sensitive mental health content using established safety and harm-reduction guidelines. Applying consistent labeling criteria using detailed annotation rubrics. Adhered to strict confidentiality and data protection protocols. Met tight turnaround deadlines while maintaining accuracy standards.