Developed the MedCalc-Bench evaluation framework within the UKGovernmentBEIS/inspect_evals repository, enabling standardized assessment of language model performance on clinical calculations extracted from free-text patient notes. The work focused on AI evaluation and machine learning, leveraging Python to implement the framework and generate an evaluation report for the gpt-4o-mini model. Enhancements included updating the README with detailed usage instructions and example workflows, as well as refining schema validation titles for clarity. The release addressed the need for reproducible benchmarking in clinical NLP tasks, with no major bugs reported or fixed during the period, reflecting a focused and well-scoped engineering effort.