AI Training Dataset Creation (Sepal AI)
I created over 100 realistic prompts for technical scenarios used in training AI models. I developed detailed binary evaluation rubrics to ensure objective, consistent labeling and dataset quality. My efforts directly improved the reliability and clarity of AI training datasets for technical subject matter. • Focused on technical content such as CRM, system admin, and troubleshooting guides • Ensured rubric reproducibility and inter-rater reliability across QA team • Used prompt engineering for scenario diversity and accuracy • Delivered measurable improvement in dataset consistency and evaluator agreement