AI Model Evaluation & Adversarial Prompting Framework
Prompt + Response Writing (SFT)Designed and authored large sets of adversarial prompts to evaluate and stress-test LLM coding reasoning. Structured captured error traces and model behavior into datasets intended for fine-tuning and RLHF feedback loops. Iteratively proposed prompt engineering improvements to measurably increase response quality. • Authored 200+ adversarial coding prompts • Verified logical soundness and factual accuracy across algorithmic topics • Logged reproducible failure cases (e.g., off-by-one, faulty recursion) • Produced structured evaluation artifacts for training feedback
2022 - 2024