Arseny Kravchenko is a seasoned ML engineer, mostly interested in computer vision problems. He has been working in machine learning since 2015, in roles ranging from individual contributor to leadership, which gives him a view of ML systems from both the keyboard and the org chart.

His specialty is turning models into systems that survive contact with production. Arseny advocates starting with a lightweight design phase and a problem-first design doc, one that splits attention evenly between the problem and the solution, and defining a system’s goals, non-goals, and assumptions before writing code. From there he builds a solution blueprint of baselines, metrics, pipeline components, and data strategy covering availability, processing, and features. He is particularly at home with edge and mobile ML constraints, where latency, frame rate, energy budgets, and Core ML force hard trade-offs, and his applied work includes projects such as photostock search and retail pricing.

Earlier in his career, Arseny was active in competitive machine learning and reached Kaggle Master level, a proving ground for exactly the kind of practical modeling judgment he now applies to production systems. He shares writing on machine learning engineering on his site, arseny.info.