Mathematics AI trainer, OutlierAI (Jan 2024–Present)
Trained and improved an AI mathematics system by structuring datasets and reducing model errors through targeted data work. Performed dataset organization and curation to increase downstream precision and completion performance. Focused on mitigating biases by revising dataset composition and annotations. • Improved model accuracy by addressing bias and dataset structure. • Optimized data annotation workflows to reduce factual/logical error rates by 15%–20%. • Curated high-quality training datasets that improved model precision/completion by 10%–15%. • Organized, cleaned, and annotated thousands of hours of raw unstructured data (text, video, audio) into supervised ML datasets.