Hayk M.

I build machine learning and data systems for problems involving prediction, risk, optimization, and automated decisions. My work covers the full process, from understanding the business problem and preparing the data to statistical modeling, deployment, and monitoring. Some examples of my work: **Credit risk modeling:** Rebuilt a bank's retail credit scoring system using ensemble ML models. Non-performing loans fell from 11% to 6%, while loan approvals increased by 10%. **Pricing and demand forecasting:** Developed models for retail businesses to improve pricing and inventory decisions. The resulting systems helped clients improve profit margins by 10–15%. **Real estate analytics:** Built data pipelines collecting and processing information from more than 20 UK property data sources, along with models for property valuation, address matching, and analytical insights. **Algorithmic trading:** Developed statistical and machine learning models, automated trading strategies, and supporting infrastructure for backtesting, execution, and risk management. My main technical stack includes Python, SQL, scikit-learn, PyTorch, XGBoost, PostgreSQL, Airflow, MLflow, Docker, and AWS. I can work on a specific modeling or statistical problem, review an existing technical solution, or take responsibility for an entire ML system from design through implementation. My approach is to understand the underlying problem before choosing a technical solution. That means questioning assumptions, selecting appropriate validation methods, and paying attention to how a model behaves outside its development environment. I value clear communication, sound statistical reasoning, and solutions whose behavior and limitations can be understood.