AI Trainer – Docent Personal AI Project
TextDeveloped and evaluated a fully on-device LLM and retrieval-augmented generation (RAG) system using structured data for persona-consistent output responses. Conducted iterative output quality tuning, temperature optimization, and hallucination mitigation using a markdown-based knowledge base. Oversaw the AI’s behavior through systematic assessment and prompt engineering in a controlled local environment. • Performed content review, output scoring, and preference data tasks for LLM output assessment. • Employed knowledge base curation and prompt evaluation for improved retrieval behavior. • Audited instances of hallucinations and guided mitigation processes. • Utilized quality evaluation instincts to ensure structured, high-fidelity training data.