Databricks Certified Machine Learning Associate

Overview of Databricks ML capabilities and MLOps best practices, including ML runtimes, AutoML for model/feature selection, Unity Catalog feature store creation/use (online vs offline), MLflow tracking and registry operations (logging, registering, tagging, promoting models), and governance considerations for model/code promotion.

Fundamentals of preparing and exploring data on Databricks: computing summary statistics, outlier removal, visualizing categorical and continuous features, comparing features, imputing missing values (mean/median/mode), encoding categorical variables (one‑hot), and applying appropriate transformations (e.g., log scale).

Core model-building tasks: selecting suitable algorithms using ML foundations, mitigating class imbalance, building training pipelines, hyperparameter tuning (including Hyperopt, grid/random/Bayesian search and parallelization), cross‑validation vs train/validation splits, and evaluating models using classification and regression metrics while managing bias–variance tradeoffs.

Approaches and practices for putting models into production: differences and tradeoffs for batch, realtime, and streaming serving; deploying custom models to endpoints; performing batch inference (e.g., with pandas); realtime endpoint deployment and querying; and streaming inference patterns using Delta Live Tables and traffic splitting between endpoints.