Prompt Engineer (Data Labeling) / Generalist
I contributed to Large Language Model improvement by creating prompts and correcting model trajectories using reinforcement learning from human feedback. My role involved developing gold standard solutions and designing automatic outcome verifiers to enhance AI decision-making and problem-solving accuracy. Additionally, I evaluated and converted real-world code challenges into high-quality training data for model tuning. • Created complex prompts requiring external tool use • Designed and applied mathematical verifiers to validate model outputs • Improved AI decision pathways by gold trajectory mapping • Transformed GitHub pull requests into actionable coding evaluation tasks