Prompt Engineer – SFT Data Labeling and Prompt Engineering
As a Prompt Engineer at Zuoyebang, I constructed SFT (Supervised Fine-Tuning) training datasets for educational large language models. I created over 50,000 high-quality instruction-response pairs to improve model performance. I also built and optimized an automated data synthesis pipeline through prompt engineering. • Developed instruction-response pairs to train AI in educational tasks • Increased data generation efficiency through automation • Enhanced model accuracy on educational benchmarks by 15% • Reduced manual annotation costs by 40%