Designing Machine Learning Systems

Why I Picked It Up / How It Read

I picked this up to upskill my data engineering skills, as I recognize that moving, maintaining, storing and serving data for regular transactional needs of backend systems are not enough. Data Engineering books/courses do touch on the analytical side of things but not deeply enough to cover the needs of systems that rely heavily on Machine Learning, but rather just simple exploratory or summary analytical needs. This is why I'm picking this book because it dives deep in how to build machine learning systems and the MLOps involved.

The book so far reads fine and has a decent organization, but what I am enjoying the most is its theoretical nature, and not leaning into exploring the technicalities of a certain algorithm or model.

What I Took Away

• The different nature of the highly iterative process surrounding serving and maintaining machine learning models makes designing these systems look closer to DevOps rather than simple Data Engineering.
• The explanation for why we might want to have data lineage and model artifacts with versioning was refreshingly clear and instantly convincing.
• I'll add more takeaways as I work through the book but I'm hooked and expect to finish it quickly.