EMMAH GITHINJI

I’m an AI Trainer and native Swahili linguist with advanced English proficiency (C2). I help teams build better machine learning systems by delivering reliable data annotation, collection, and quality evaluation that directly improves model outcomes.

In my current AWS re/Start Program, I engineered RLHF workflows and ran pairwise comparisons that helped uncover a 20% model failure rate for FinOps anomaly detection. I also built synthetic data pipelines with Python and LangChain to automate prompt evaluation across 500+ test cases, cutting manual review time by 40%.

I’ve created golden datasets with 98% alignment to quality rubrics by collecting and pre-processing raw data with strong attention to measurement and consistency. Previously, as a Data Annotation Specialist (freelance/contract), I transcribed sensitive audio and escalated critical cases with 100% accuracy, ensuring safety-critical information reached care coordinators.

I also digitized and structured data (like PDF invoices) to reduce retrieval time by 50%, and completed web research across 200+ companies with 98% accuracy through rigorous data validation. Across legal and general domains, I’ve transcribed, translated, and annotated audio/video content—reducing legal terminology error rates from 15% to 3%—and I apply the same quality discipline to translation and transcription work.