Human Data Engineer (Contract)
I engineered adversarial C++ and Java tests to challenge and evaluate AI code generation models. My responsibilities included generating counter-examples and verifying model output accuracy to identify and reduce hallucinations in code responses. I ensured all testing adhered to strict false positive and negative rate guidelines specified by the team. • Evaluated algorithmic output and code completions produced by AI models. • Developed and supplied feedback on edge-case scenarios to improve model robustness. • Documented results and validation processes to ensure reproducible code testing outcomes. • Contributed to lowering model hallucinations by providing diverse adversarial examples.