MLOps

About the MLOps certifications here

MLOps is the work of getting machine learning models out of notebooks and keeping them useful in production: tracking experiments, registering and versioning models, serving them, watching them for drift and retraining them when they degrade. The certifications in this group test that work on a real platform rather than machine learning theory alone.

Right now both certs here come from Databricks, whose program is organized by role and level. Each role, such as data engineer or machine learning engineer, has an Associate exam for people doing everyday tasks on the platform and, for some roles, a Professional exam for people running that work in production at scale. Databricks also offers free fundamentals badges for beginners, but they are not covered here.

• Databricks Certified Machine Learning Associate covers the day-to-day ML workflow: data preparation, training and tuning, MLflow tracking and autologging, AutoML, feature tables and the model registry in Unity Catalog, and batch or real-time scoring.
• Databricks Certified Machine Learning Professional is the senior step. It is about distributed training and tuning, moving models between environments, testing, Lakehouse Monitoring for drift, automated retraining and serving strategies such as canary rollouts.

Both are proctored multiple-choice exams, and Databricks asks you to recertify by taking the current version of the exam again.

Which exam to start with

Start with the Machine Learning Associate, even if you have years of ML experience. Its questions are as much about Databricks itself as about modelling, and the Professional exam assumes you already know MLflow tracking, the Unity Catalog registry and basic Model Serving. If the Associate tests come easily, move on to the Professional. If your work is mostly with LLMs, retrieval and agents rather than classic models, look at the Generative AI Engineer Associate in the GenAI group instead.

How the practice questions here work

Each cert has several tests of 10 questions that mix the exam's sections in roughly its proportions, with topics labelled finely enough to show which areas you keep missing. Every question explains the right answer and each wrong option, links to the Databricks or MLflow documentation behind it, and has its own page. There are no full-length practice exams for these certs yet. Your scores stay in your browser unless you sign in to sync them.

Databricks renames features often, so the questions follow the current documentation: Unity Catalog versions of the model registry and feature tables, not the older workspace ones.

These questions are original, written from public documentation. This site is not affiliated with Databricks.