AI Trainer / AI Data Specialist (Project Aether & Project Hedgehog), Handshake AI & Outlier
As an AI Trainer and Data Specialist for Handshake AI & Outlier, I evaluated, annotated, and rated AI-generated textual responses, supporting large language model (LLM) development. I compared outputs, ranked responses, and identified reasoning errors and hallucinations across complex, multi-step prompts to enhance accuracy and safety. This role required strict adherence to structured guidelines and comprehensive feedback for ongoing model improvement. • Conducted detailed prompt engineering, response ranking, and annotation for LLM datasets. • Provided structured, written justifications to improve training data quality. • Identified AI hallucinations, logical inconsistencies, and instruction-following failures. • Performed scenario-based and comparative response assessments for model refinement.