Production AI and data-engineering systems, plus independent builds — led by a full-stack RAG application that shipped real business impact.
An OpenAI-powered system that generates structured attributes for 42,000+ product records.
Replaced 12+ months of projected manual research with a 3-week run; cut the data-entry backlog by over 95%.
A production ML pipeline that resolves duplicate customer records across CRM and ERP systems.
Deduplicated 20,000+ records at 90% accuracy, eliminating months of manual reconciliation.
ETL, data-warehouse, and mapping work powering a mission-critical 2-year ERP-to-NetSuite migration.
Built SQL ETL with regex + multi-currency logic and an auto-updating mapping ETL feeding executive Power BI dashboards.
A full-stack prompt-engineering platform to create, test, and version LLM prompts with real-time evaluation.
Automated quality scoring across prompt structure, clarity, and length.
A PyTorch image super-resolution model (CNN) that upscales low-resolution images.
Custom loss functions and training pipelines aimed at practical computer-vision use.