Designing Machine Learning Systems

The ML lifecycle seen from an engineer’s seat: data, features, deployment, monitoring, drift. Useful less for its models than for its frame — it forces the question “what will this look like in production six months from now” while the task is still being defined.