Hari Prasath S

•Reviewed video clips to assess caption accuracy and alignment with visual content.
•Identified and categorized caption mismatches, missing events, and descriptive inconsistencies.
•Provided detailed reports on annotation errors and discrepancies to support quality improvement.
•Followed annotation guidelines to maintain consistency and accuracy in AI training data.

•Evaluate and review AI-generated responses in Tamil for accuracy, relevance, fluency, and cultural appropriateness.
•Perform Reinforcement Learning from Human Feedback (RLHF) tasks by comparing, ranking, and assessing AI model outputs.
•Apply project-specific guidelines to ensure high-quality annotations and consistent evaluations.
•Identify factual, linguistic, and contextual issues in AI respon

Working as a generalist LLM trainer and data annotator on AI training data pipelines.
• Rate and evaluate AI-generated responses for quality, accuracy, relevance, safety, and instruction-following across diverse task types
• Annotate text data, including classification, named entity recognition, and instruction-response pairs for large language model training
• Deliver structured RLHF (Reinforc

MLOps internship focused on building training datasets and deploying YOLO-based object detection models for industrial safety monitoring.
• Performed polygon annotation and bounding box labeling on industrial images using LabelMe and LabelImg
• Built and trained three YOLO (v5/v8) object detection models from scratch for conveyor belt obstacle detection
• Handled full dataset pipeline: collect

Oct 2025

Feb 2025

Sep 2024