DBX-MLP Databricks Machine Learning Professional
- Vendor
- Databricks
- Level
- Professional
- Questions
- 59
- Time
- 120 min
- Domains
- 3
- In the bank
- 355
Practice 355 questionsTake a timed mock examDay-by-day study plan
Guides for this exam
- 8 Mistakes That Repeat Across Every Databricks Exam, and How to Spot Each One in the Question · 한국어
- DBX-MLP Study Guide: Study the Hand-offs, Because That Is Where Production ML Breaks · 한국어
The exam at a glance
| Level | Professional |
|---|---|
| Questions | 59 |
| Time | 120 minutes (about 2 minutes per question) |
| Passing mark | Not published by the vendor |
Domains and weights
| Domain | Weight | Practice questions here |
|---|---|---|
| D1 · Model Development | 44% | 158 |
| D2 · MLOps | 44% | 155 |
| D3 · Model Deployment | 12% | 42 |
D1 · Model Development 44%
Professional model development questions mix Spark ML, distributed training and custom pyfunc models at scale. Read the data shape and where training runs. Strong options use the API that fits the workload, batch or stream, single node or many; weaker ones apply an estimator or library where it cannot run.
What to watch for
- Spark ML estimators fit on static DataFrames; a fitted PipelineModel can then score a stream.
- String labels need StringIndexer before a classifier, and IndexToString maps predictions back.
- Multi-node GPU errors such as NCCL failures usually point to networking or environment, not to the model code.
Tasks in the official exam guide (22)
- 1.1 Identify when SparkML is recommended based on the data, model and use case requirements
- 1.2 Construct an ML pipeline using SparkML
- 1.3 Apply the appropriate estimator and/or transformer for a use case
- 1.4 Tune a SparkML model using MLlib
- 1.5 Evaluate a SparkML model
- 1.6 Score a SparkML model for a batch or streaming use case
- 1.7 Select a SparkML or single-node model for batch, real-time or streaming inference
- 1.8 Scale distributed training pipelines using SparkML and pandas Function APIs/UDFs
- 1.9 Perform distributed hyperparameter tuning with Optuna and integrate it with MLflow
- 1.10 Perform distributed hyperparameter tuning with Ray
- 1.11 Evaluate trade-offs between vertical and horizontal scaling for ML workloads
- 1.12 Select parallelization strategies (model or data parallelism) for large-scale training
- 1.13 Compare Ray and Spark for distributing ML training
- 1.14 Use the pandas Function API to parallelize group-specific model training and inference
- 1.15 Use nested runs in MLflow to track complex experiments
- 1.16 Log custom metrics, parameters and artifacts programmatically in MLflow
- 1.17 Create custom model objects using real-time feature engineering
- 1.18 Ensure point-in-time correctness in feature lookups to prevent data leakage
- 1.19 Build automated feature computation pipelines with the FeatureEngineering Client
- 1.20 Configure online tables for low-latency applications using the Databricks SDK
- 1.21 Design scalable streaming ingestion and processing to generate features in real time
- 1.22 Develop on-demand features with feature serving for consistent use in training and production
D2 · MLOps 44%
MLOps questions cover the path from training to production: MLOps Stacks, Asset Bundles, the Unity Catalog registry and monitoring. The deciding detail is usually which environment holds which permissions and when metrics are compared. Strong options compare models on the same current data and keep production locked down.
What to watch for
- Compare champion and challenger on the same fresh evaluation data, not on metrics stored when each was trained.
- Asset Bundles define jobs, endpoints and their traffic splits as code per target environment.
- Data scientists can analyze production through read-only inference tables and monitoring, without write access.
Tasks in the official exam guide (20)
- 2.1 Describe and implement the model lifecycle pipeline components that manage environment transitions in the deploy-code strategy
- 2.2 Map Databricks features to activities of the model lifecycle process
- 2.3 Implement unit tests for notebook functions
- 2.4 Identify the unit and integration tests used at each environment stage (dev, test, prod)
- 2.5 Design an integration test for ML pipelines (features, training, evaluation, deployment, inference)
- 2.6 Compare approaches for organizing functions and unit tests
- 2.7 Design and implement scalable Databricks environments for ML projects using best practices
- 2.8 Define Databricks ML assets with bundles: serving endpoints, MLflow experiments, registered models
- 2.9 Implement automated retraining triggered by data drift or performance degradation alerts
- 2.10 Develop a strategy for selecting top-performing models during automated retraining
- 2.11 Apply statistical tests from the drift metrics table to detect drift in numerical and categorical data
- 2.12 Identify the table type and monitoring feature that fit a use case, and explain why
- 2.13 Build a monitor for a snapshot, time series or inference table
- 2.14 Identify key components of monitoring pipelines (logging, drift, performance, health)
- 2.15 Design alerting that notifies stakeholders when drift metrics exceed thresholds
- 2.16 Detect drift by comparing current data to a baseline or successive time windows
- 2.17 Evaluate model performance trends over time using an inference table
- 2.18 Define custom metrics in monitoring metrics tables
- 2.19 Evaluate metrics at different data granularities and feature slices
- 2.20 Monitor endpoint health: latency, request rate, error rate, CPU and memory
D3 · Model Deployment 12%
Deployment questions focus on Model Serving endpoints: rollouts, traffic splits and checks before going live. Read the traffic volume and what the update changes. Strong options pick a rollout that still gives a meaningful signal and validate the environment first; weaker ones run a canary too small to measure.
What to watch for
- A configuration update keeps the old version serving until the new one is ready.
- At low traffic a small canary percentage yields too few requests; use a larger share or a shadow test.
- Validate a model's environment before serving, for example with mlflow.models.predict in a clean environment.
Tasks in the official exam guide (5)
- 3.1 Compare deployment strategies (blue-green, canary) and their suitability for high-traffic applications
- 3.2 Implement a model rollout strategy with Model Serving
- 3.3 Register a custom PyFunc model and log custom artifacts in Unity Catalog
- 3.4 Query custom models via REST API or the MLflow Deployments SDK
- 3.5 Deploy custom model objects with the MLflow Deployments SDK, REST API or UI
A study order that works
The 19-day study plan for DBX-MLP takes the domains in this order. Each day has a short reading, key terms and a quiz from the bank.
- Days 1–8: D1 · Model Development
- Days 9–16: D2 · MLOps
- Days 17–18: D3 · Model Deployment
- Day 19: Final Review: Decision Cheat-Sheet Across the Three Domains
Recent changes to these questions
- : 1 answer choice set changed, 1 answer key corrected, 1 explanation rewritten
- : 60 questions added
- : 1 answer choice set changed, 1 answer key corrected, 1 explanation rewritten
- : Exam added to the bank (295 questions added)
About these questions
Every question is original, written from the public exam guide and checked against the vendor's own documentation. None comes from a real exam. Each one has an explanation of why the answer is right and why the other options are wrong. How the questions are written.