Dmytro Vykhovanets

I’m a Senior Data Scientist and AI Engineer with 4+ years delivering production machine learning and generative-AI systems for enterprise banking, fintech, industrial supply, and U.S. healthcare clients. I start from ambiguous business needs and translate them into measurable solutions I can ship and defend in production.

My strongest work sits at the intersection of classical ML and modern LLMs: retrieval-augmented systems, agentic workflows, schema and entity mapping, NLP, and document AI. I design data pipelines, train and fine-tune models, and build LLM agents that turn retrieved context into operational decisions.

Most recently, I built AI data-ingestion and cataloging capabilities for a major U.S. building-materials supplier—normalizing hundreds of manufacturer formats into a single 250+ column schema with a schema-mapping model (85% accuracy@5) and human review for low-confidence cases. I also created a hierarchical product-categorization system (83% accuracy through level 3) and extended it with OCR plus a vision LLM to extract structured attributes from spec-sheet PDFs, removing manual bottlenecks.

Earlier, I led work across RAG assistants and predictive systems: a board-member RAG analysis support tool using AWS Bedrock and Milvus with hybrid entity recognition, an adaptive learning recommendation engine improved through hypothesis development and A/B testing, and an end-to-end documentation autogenerator integrated into CI/CD. Across contracts, I’ve delivered an investment-analytics RAG chatbot, validation-and-recalculation pipelines for sensitive financial data, an Email-to-Jira converter (87% accuracy), and a natural-language-to-SQL engine that earned a 4.7/5 marketplace rating.