AI / ML engineer · multi-agent systems
I build systems that learn to coordinate under pressure.
Reinforcement learning, simulation, and applied ML — with a bias toward the messy conditions where systems actually break. Below is selected work.
Experience
I work on the part of machine learning that has to survive contact with the real world.
I design, train, and ship end-to-end GenAI systems — RAG pipelines, transformer NLP, agentic workflows, knowledge-graph retrieval, and multi-agent reinforcement learning — and I take them all the way from prototype to production.
I hold a Master's in Information and Data Science from UC Berkeley, and I have fourteen years building software: the last several in AI and machine learning, the earlier ones leading test automation for cloud platforms at scale.
That combination is why I tend to own the entire lifecycle myself — data engineering, modeling, fine-tuning, MLOps — and why I care more about how a system degrades than how it scores on a good day. The work below is where that shows.
Education
Where it came from
M.S. Information & Data Science (MIDS)
Expected 2026UC Berkeley, School of Information
Machine Learning · NLP with Deep Learning · Statistics · Data Engineering · Data Visualization · Research Design
B.A. Economics
2008National Technical University, Kharkiv, Ukraine
Certifications
- Multi-AI-Agent Systems with CrewAI — DeepLearning.AI
- AI Agents with LangChain & LangGraph — Udacity
- Build AI Apps with MCP Server — DeepLearning.AI
- AI Strategy — Augment.org
- IBM Python for Data Science & AI
Toolkit
Skills
Machine learning
GenAI & LLMs
NLP
Languages & frameworks
Data, cloud & MLOps
Selected work
Projects
Each case study covers the problem, the approach, and what the numbers actually showed — including where they did not hold.
Contact
Working on something that has to coordinate?
Open to research collaborations and applied ML work in robotics, simulation, and multi-agent systems.