At DylogAI, I engineered a modular classification benchmarking pipeline that brought embeddings, local models, and LLM APIs under a shared I/O contract. It accelerated cross-model evaluation by 35% and reduced redundant API token consumption by 22%.
At Tennessee Valley Authority, I'm implementing a Power BI Data Platforms Health Scorecard and forecasting Azure and Gateway resource thresholds from utilization trends. At Purdue University, I refactored data conversion pipelines and developed validation suites across 50,000+ records, reducing pipeline error rates by 30%.
