At Zinkworks, I take forecasting and product-discovery features for an AI-assisted commerce analytics product from prototype through production. I designed a FastAPI inference layer and cut the forecasting API’s p95 response time from 820 ms to 510 ms.
Previously, I built Python services and data pipelines at Atobi, Accenture Baltics, and Dreamix. My work included recommendation models, production data workflows, and deployment systems using Python, AWS, and machine-learning tools.

