Databricks ML Associate - Study Plan¶

5-Week Study Schedule¶

Week 1: ML Workloads on Databricks (Domain 1)¶

Day 1-2: Databricks ML Runtime and MLflow Tracking¶

• Understand ML Runtime libraries and GPU options
• Learn MLflow tracking: log_param, log_metric, log_artifact, log_model
• Practice MLflow autologging with sklearn and XGBoost
• Understand experiments, runs, and nested runs
• Review Notes: notes/01-ml-workloads.md
• Read: MLflow Tracking

Day 3-4: Model Registry and Feature Store¶

• Learn Unity Catalog Model Registry vs workspace registry
• Practice registering models and managing versions
• Understand model aliases (champion, challenger)
• Learn Feature Store: create_table, write_table, create_training_set
• Read: Model Registry
• Read: Feature Store

Day 5-6: AutoML¶

• Run AutoML for classification, regression, and forecasting
• Review generated notebooks and model leaderboard
• Understand AutoML capabilities and limitations
• Learn when to use AutoML vs custom training
• Read: AutoML

Day 7: Week 1 Review¶

• Review all MLflow APIs and patterns
• Quiz yourself on Feature Store workflow
• Practice Model Registry operations

Week 2: ML Workflow (Domain 2)¶

Day 8-9: Data Preparation and Feature Engineering¶

• Practice train/test splits and stratified splitting
• Learn feature scaling: StandardScaler, MinMaxScaler
• Practice categorical encoding: one-hot, label encoding
• Understand missing value strategies
• Review Notes: notes/02-ml-workflow.md

Day 10-11: Hyperparameter Tuning with Hyperopt¶

• Learn Hyperopt: fmin, search spaces, tpe.suggest
• Practice SparkTrials for distributed tuning
• Understand search space functions: hp.choice, hp.uniform, hp.loguniform
• Integrate Hyperopt with MLflow tracking
• Read: Hyperopt

Day 12-13: Model Evaluation¶

• Master classification metrics: accuracy, precision, recall, F1, AUC-ROC
• Master regression metrics: RMSE, MAE, R-squared
• Understand overfitting vs underfitting diagnosis
• Practice cross-validation techniques
• Read: Model Evaluation

Day 14: Week 2 Review¶

• Compare Hyperopt search strategies (grid, random, Bayesian)
• Review when to use which evaluation metric
• Practice identifying overfitting from train/test metrics

Week 3: Spark ML and Deep Learning (Domains 3-4)¶

Day 15-16: Spark MLlib Pipeline API¶

• Learn Transformers, Estimators, and Pipelines
• Practice VectorAssembler, StringIndexer, OneHotEncoder
• Build complete ML pipelines with preprocessing and training
• Review Notes: notes/03-spark-ml.md
• Read: Pipeline API

Day 17-18: Model Selection with Spark¶

• Practice CrossValidator with ParamGridBuilder
• Compare CrossValidator vs TrainValidationSplit
• Learn Spark ML evaluators for different problem types
• Save and load pipeline models
• Read: Model Tuning

Day 19-20: Deep Learning on Databricks¶

• Understand GPU runtime and instance types
• Learn transfer learning concepts (freeze, fine-tune)
• Understand distributed training: TorchDistributor, Horovod
• Practice MLflow integration with TensorFlow and PyTorch
• Review Notes: notes/04-deep-learning.md
• Read: Deep Learning

Day 21: Week 3 Review¶

• Build a complete Spark ML pipeline from scratch
• Review distributed training approaches
• Quiz yourself on Pipeline API components

Week 4: Scaling ML and Integration (Domain 5)¶

Day 22-23: Pandas UDFs and Pandas API on Spark¶

• Practice pandas UDF decorator patterns
• Understand Series-to-Series vs iterator UDFs
• Use Pandas API on Spark for scalable data exploration
• Review Notes: notes/05-scaling-ml.md
• Read: Pandas UDFs

Day 24-25: Inference Patterns¶

• Practice batch inference with mlflow.pyfunc.spark_udf
• Understand streaming inference with MLflow models
• Learn Model Serving endpoint concepts
• Compare batch vs streaming vs real-time inference
• Read: Batch Inference

Day 26-28: Cross-Domain Integration¶

• Design end-to-end ML workflows: data prep to inference
• Combine Feature Store with MLflow tracking
• Practice full pipeline: features, training, registry, inference
• Review all notes files for connections between domains

Week 5: Exam Preparation¶

Day 29-30: Comprehensive Review¶

• Re-read all five notes files and fact sheet
• Focus on highest-weight domains: ML Workloads (29%) and ML Workflow (29%)
• Create a summary of key API patterns
• Review key differentiators

Day 31-32: Practice Scenarios¶

• Work through scenarios.md exam-style questions
• Take available practice exams
• Review incorrect answers and understand reasoning
• Focus on weak areas

Day 33-35: Final Preparation and Exam¶

• Light review of key concepts
• Review exam tactics and time management strategy
• Take the exam with confidence

Study Tips¶

Time Allocation by Domain Weight¶

Recommended Daily Schedule¶

• 30 min reading documentation or notes
• 30 min hands-on practice with code examples
• 15 min reviewing API patterns and key concepts
• Total: ~1.25 hours/day