MLOps Zoomcamp: a free 9-week course repository, and what it actually contains

The problem this repository solves is sequencing, not software

Most MLOps material arrives as documentation for a single tool. You can read the MLflow docs, the Evidently docs and the Terraform docs and still have no ordered path through them. This repository is that ordering. It is a curriculum of six modules plus a final project, and the running example across the course is the NY Taxi dataset, so the data does not change under you while the tooling does. The stated audience is data scientists, ML engineers and software engineers who want to put models into production and operate them. The prerequisites are explicit: Python, Docker, command line basics, machine learning (the README points at ML Zoomcamp), and one or more years of programming experience. That last requirement matters. This is not an introduction to programming with a machine learning flavour. If you have never containerised anything, module 4 will be rough.

What each of the six modules actually covers

Module 1 sets up the environment and introduces the MLOps maturity model. Module 2 is experiment tracking and model management with MLflow: saving and loading models, and the model registry. Module 3 covers orchestration and ML pipelines. Module 4 splits deployment into online (web and streaming) and offline (batch), with Flask for the web service, AWS Kinesis and Lambda for streaming, and batch scoring for offline. Module 5 is monitoring, and it is the most tool-dense module: a web service path through Prometheus, Evidently and Grafana, and a batch job path through Prefect, MongoDB and Evidently. Module 6 covers unit and integration testing, linting, formatting, pre-commit hooks, CI/CD with GitHub Actions, and infrastructure as code with Terraform. The final project in 07-project is meant to integrate all of it into one pipeline. Read that list as a scope statement. It is breadth over depth: you will touch MLflow, Flask, Kinesis, Lambda, Prometheus, Evidently, Grafana, Prefect, MongoDB, GitHub Actions and Terraform in nine weeks. No single tool gets the treatment a dedicated course would give it.

How the course is delivered, and what changed for 2026

There are two modes. Live cohorts run with pre-recorded lectures, graded homework, a leaderboard, peer review and a certificate, all free. Self-paced study uses the same pre-recorded lectures, and homework is available but not scored, with no leaderboard, no peer review and no certificate. The README carries a note stating that no live cohort is planned for 2026 and that the course is fully available for self-paced study, with a registration link to be notified if a live cohort returns. This is the single most consequential fact about the project right now. The certificate image and the certificate guide still sit in the README, but they describe the live-cohort path. If a certificate is your reason for enrolling, verify the cohort status before you start, because the self-paced route does not produce one. The self-paced steps are listed plainly: follow the materials on GitHub, ask questions in Slack, do the homework as self-check, and build a project for your portfolio.

Getting started: clone the repo and work module by module

The materials live in this repository, with one directory per module (01-intro through 06-best-practices) and 07-project for the final project. The README's self-paced instructions are to follow the materials on GitHub, ask questions in Slack, and complete the homework and a project. Video lectures are on the linked YouTube playlist, and deadlines and homework submission run through courses.datatalks.club. Environment setup happens in module 1, so the concrete commands and configuration keys you will type come from the module notebooks and instructions rather than from the top-level README, which is a course index rather than a setup guide. That is a real friction point: the README tells you what the course contains and where to register, but the installation steps for MLflow, Prefect, Evidently and the rest are inside the module folders. Budget time for reading those before you assume a tool is broken. Support runs through the #course-mlops-zoomcamp channel on DataTalks.Club Slack, with a separate Telegram channel for announcements and a FAQ document for everything else.

Where the format breaks down

Self-paced study removes the two mechanisms that make a cohort work: graded feedback and peer review. Homework is available but not scored, so nothing tells you whether your MLflow tracking server is configured the way the instructors intended. You can compare your output against the module notebooks, and that is the extent of the correction loop. The second limitation is tool churn. The course pins specific services: AWS Kinesis and Lambda for streaming, MongoDB for batch monitoring, Prefect for orchestration. If your employer runs Kafka and Airflow, module 5's batch monitoring path teaches you the shape of the problem with tools you will not use. The transferable part is the monitoring design, not the MongoDB queries. Third, the breadth means the final project is where the integration actually happens, and it is self-directed. A learner who follows modules 1 through 6 and skips the project has a set of disconnected exercises. The repository also lists no releases, so there is no versioned snapshot of the curriculum to pin against; the main branch is the course.

Alternatives, and the difference in approach

The obvious comparison is ML Zoomcamp, the same organisation's machine learning course, which the README lists as a prerequisite. The split is deliberate: ML Zoomcamp teaches modelling, MLOps Zoomcamp assumes you can already model and teaches what happens after the notebook. If you cannot yet train and evaluate a model, MLOps Zoomcamp is the wrong starting point, and the prerequisite list says so. The other comparison is a vendor's own tutorial track, such as the MLflow or Evidently documentation quickstarts. Those go deeper on one tool and assume you already know where it fits. This repository goes the other way: shallow on each tool, explicit about the order and the connections between them. The trade is depth for a map. If you already know which tool you need and just want its API, the vendor docs are faster. If you know the tools exist but not how a deployment, a monitoring stack and a CI pipeline relate to each other, the module sequence is the thing you cannot get from any single project's documentation.

Maintenance, licensing and what the material does not state

The repository is not archived, the default branch is main, and the last push recorded is 2026-09-10, so the curriculum is being kept current. There are no releases retrieved, which means there is no changelog to diff between cohorts; if you want to know what changed since you last looked, you compare the module folders yourself. The license is listed as unknown. That is worth flagging without over-reading it: a course repository typically contains notebooks, instructions and slides rather than a distributable library, so the absence of a license file may simply reflect that the project never needed one. It still means you should not assume you can reuse the material in your own training product without asking. The README does not state a license, a contribution process beyond a PRs-welcome badge, or a support commitment. Sponsorship enquiries go to alexey@datatalks.club, which is the only commercial contact listed.

Editorial conclusion

Adopt it if you already write Python, use Docker and the command line, and want a guided path through MLflow, Flask deployment, Evidently monitoring and Terraform. Skip it if you need graded feedback or a certificate: the README states no live cohort is planned for 2026, and self-paced homework is available but not scored. Before committing nine weeks, open the 01-intro folder, confirm the NY Taxi dataset and environment setup match the machine you will use, and check the FAQ at datatalks.club/faq/mlops-zoomcamp.html for the current cohort status.

Official sources

1. DataTalksClub/mlops-zoomcamp on GitHub
2. Issues
3. Project website
4. README