Selected work
A sample of the technical work I enjoy most — evaluating AI systems, removing manual steps from workflows, and making data easier to read.
A rubric-driven workflow for scoring large language model responses on accuracy, reasoning quality, and guideline adherence. Includes prompt templates, a scoring schema, and reporting that surfaces where a model drifts.
A set of Python and SQL scripts that replace repetitive reporting and data-entry work. Pulls from internal sources, validates records, and publishes clean summaries on a schedule.
An exploratory dashboard turning raw survey and performance data into readable trends. Combines statistical summaries with charts designed for non-technical stakeholders.