Machine Learning Software Engineering
Machine learning is the next generation of software engineering, and this means we need a start a cultural shift towards data scientists becoming active and productive participants in the software engineering process. A key part of this is reducing the friction for data scientists to think about coding “non-interactively” and building models and behavioural tests that can run as part of a DevOps pipeline.
Praneet Solanki from the Azure CAT team has been building out a reference architecture for this pattern “MLOps”, which he presented at our internal MLADS conference with several of his colleagues. Praneet will walk through the architecture on a follow-up video but here we discuss the overarching issues. Galiya Warrier is from the CSE team and has been involved in ML DevOps architectures with some of our large customers.