Databricks Machine Learning Associate: training and operations

Objectives and progression

Independent preparation aligned with the four domains of the Machine Learning Associate guide still linked by the official page in October 2026. Learn to review AutoML trials, preserve temporal feature meaning, track MLflow experiments, and distinguish model promotion from actual deployment. Lessons connect data preparation, pipelines, parameter search, and metrics to operational use. Seven fictional cases include pressure to promote an invalid metric, stale inputs overriding lookups, training budget, review capacity, and a removed endpoint identity. Includes 28 primary sources, progress, and an internal 28-decision assessment in 60 minutes. Explains differences between blueprint terminology and current runtime features; this initial coverage is not a full mock or a substitute for supervised practice.

Audience: Data, Python, APS, and project professionals preparing and operating machine learning models.

Prerequisites: Python, SQL, and statistics and ML fundamentals. The exam has no prerequisites; practical experience is recommended.

340 estimated study minutes

• Inspect experiments and validate dependencies before accepting a metric.
• Prevent temporal leakage and silent differences between training and inference.
• Connect results to versions and distinguish registration from deployment.
• Prepare features without using information from the held-out set.
• Compare alternatives with controlled cost and validation.
• Interpret errors in the units and context of the operational decision.
• Choose consumption mode and control identity, version, and traffic.

Modules

1. Environment, AutoML, and reproducibility
2. Temporal features and consistency
3. MLflow, registry, and promotion
4. Data, preprocessing, and validation
5. Training, pipelines, and parameter search
6. Metrics, decisions, and generalization
7. Inference, serving, and recovery

Continue learning

Data / AI

• Python
• SQL
• Databricks Certified Data Engineer Associate
• AWS Certified Machine Learning Engineer - Associate
• AWS Certified AI Practitioner

References and version

2025-03-01

• Databricks Certified Machine Learning Associate · 2026-10-01
• Machine Learning Associate exam guide effective 1 March 2025 · 2026-10-01
• AutoML trial notebooks and runtime support · 2026-10-01
• Hyperparameter tuning and Hyperopt runtime changes · 2026-10-01
• Hyperopt fmin Trials and SparkTrials concepts · 2026-10-01
• MLflow 3 installation and migration · 2026-10-01
• Model lifecycle in Unity Catalog · 2026-10-01
• Feature engineering and serving overview · 2026-10-01
• Feature tables in Unity Catalog · 2026-10-01
• Point-in-time feature lookups · 2026-10-01
• Train and score with feature metadata · 2026-10-01
• Track experiments and training runs · 2026-10-01
• Deploy models using Model Serving · 2026-10-01
• Create custom model serving endpoints and identity · 2026-10-01
• Traffic routing across served models · 2026-10-01
• Delta Lake streaming reads and writes · 2026-10-01
• Lakeflow Spark Declarative Pipelines · 2026-10-01
• Prevent inconsistent preprocessing and data leakage · 2026-10-01
• Imputation of missing values · 2026-10-01
• Cross-validation and time series splits · 2026-10-01
• OneHotEncoder unknown categories · 2026-10-01
• Precision and recall interpretation · 2026-10-01
• Mean absolute error · 2026-10-01
• Model evaluation metrics · 2026-10-01
• GridSearchCV folds and refit · 2026-10-01
• Transforming regression targets · 2026-10-01
• Spark ML Estimators Transformers and Pipelines · 2026-10-01
• PySpark DataFrame summary statistics · 2026-10-01