At Donkit, I built and delivered an MVP of a RAG-based agent that automatically designs retrieval pipelines for client datasets. I also developed synthetic data and benchmarking pipelines, evaluation methods, and workflows for agent performance analysis and self-improvement.
At MarineTech, I co-authored a neural network architecture for predicting ocean currents and built statistical models for medium-range weather forecasting. I trained and deployed models as web services and developed a computer vision service for walrus detection in aerial photographs.
At Mind&Machine, I developed workforce management solutions for retail businesses and supported product integration into three retail chains. My work included time-series analysis, data pipelines, and machine-learning configuration and customization.
At SeeTree, I created a proof-of-concept module to calculate daily radiation exposure for trees and analyzed orange plantation health using NDVI and multispectral data. I also found data correlations that helped track trees receiving suboptimal irrigation.
