● Built an ML - based demand forecasting model for a client in the alcohol industry using the Snowflake data ecosystem. ● Designed a multi - model framework to handle diverse time series across tenants, improving forecast accuracy with a final WAPE below 25%.

● Integrated ML and deep learning models into production workflows, ensuring their stability, relevance, and continuous improvement. ● Used an ensemble of CatBoost and neural network models combined through Bayesian averaging to forecast the minimum methanol dosage, reducing annual consumption by 13% YoY. ● Built an ML classification pipeline to optimize corrosion inhibitor usage, applying gradient boosting with a monotonic constraint on the “dose” feature and confidence intervals, which reduced inhibitor overuse by 30% YoY while keeping corrosion rates below 0.01 mm/year. ● Developed an AI-powered document search system based on a RAG (Retrieval-Augmented Generation) architecture, implementing hybrid search with Bi-Encoder and Cross-Encoder models combined with LLM integration via API.

● Data collection and data processing. ● Modeling and communication of results.