AI Data Trainer & Knowledge Base Lead
Led the creation and iteration of SFT data annotation standards from scratch to support LLM-based enterprise knowledge base development. Designed and managed a three-level annotation and quality acceptance workflow, accumulating 8,000 high-quality fine-tuning instruction data entries to optimize model reasoning and compliance. Built a Golden Dataset and model evaluation system encompassing complex intent classification and prompt-based data generation. • Developed annotation SOPs and badcase definitions for over 50 business intents • Trained and managed annotation team, created metadata standards for semantic integrity • Led prompt engineering and prompt+response generation for SFT data • Conducted evaluation, quality inspection, and continual model improvement through detailed case reviews