DataTalksClub/machine-learning-zoomcamp

What It Is

Machine Learning Zoomcamp is a free, practical course that teaches people how to build, evaluate, and deploy machine learning systems. It's not a theory class — the focus is on applying machine learning to real problems and turning trained models into services other applications can use.

The course follows the full path from a machine learning problem to a production service: frame the problem → prepare the data → train and evaluate the model → expose it through an API → package it → deploy it.

Who It's For

You don't need prior machine learning experience. The prerequisites are:

• At least one year of programming experience
• Comfort using the command line
• A laptop with an internet connection

Helpful (but not required): some Python familiarity, basic Git/GitHub experience, and experience working with data in notebooks or scripts. Not required: any ML training experience, cloud or Kubernetes knowledge, or a powerful computer with a GPU (deep learning modules use cloud resources).

The course is especially relevant for software engineers, data analysts, data engineers, and technical students. It's not a good fit if you have no programming experience, want a non-technical intro to AI, or want advanced ML research topics.

How the Course Works

The full path uses:

• Python, NumPy, pandas, scikit-learn — for data handling and classic ML models
• TensorFlow, PyTorch, Keras — for deep learning
• FastAPI, Docker, Kubernetes, AWS Lambda — for deployment and serving models

You can take it two ways:

	Live Cohort	Self-Paced
Start	September 14, 2026	Anytime
Cost	Free	Free
Lectures	Pre-recorded	Pre-recorded
Homework	Graded	Available, not scored
Leaderboard	Yes	No
Peer review	Yes	No
Certificate	Eligible	Not available

"Live cohort" doesn't mean mandatory live classes — lectures are pre-recorded. It means shared deadlines, scored homework, leaderboard participation, peer review, community support, and certificate eligibility.

What You'll Learn (Syllabus)

1. Introduction to ML — What ML is, when to use it, structuring projects with the CRISP-DM framework, supervised learning basics.
2. Regression — Build a car-price prediction model: linear regression from scratch and with scikit-learn, feature creation/selection, regularization, model validation.
3. Classification — Build a customer-churn prediction system with logistic regression: encoding categorical variables, feature importance, interpreting predictions.
4. Evaluation Metrics — Accuracy, precision, recall, F1 score, confusion matrices, ROC curves and AUC, cross-validation, handling imbalanced classes.
5. Deploying ML Models — Serialize trained models, expose predictions through a FastAPI service, package with Docker, deploy to a cloud platform.
6. Decision Trees & Ensemble Learning — Decision trees, random forests, gradient boosting with XGBoost, hyperparameter tuning, feature importance.
7. Midterm Project — Apply modules 1–6 to an end-to-end problem of your choice.
8. Neural Networks & Deep Learning — Neural network components, PyTorch/TensorFlow/Keras, convolutional neural networks, transfer learning, image classification.
9. Serverless Deep Learning — Deploy models with AWS Lambda and API Gateway (scikit-learn, TensorFlow, PyTorch models without managing a persistent server).
10. Kubernetes & TensorFlow Serving — Core Kubernetes concepts, serving models with TensorFlow Serving, deploying and scaling model services.
11. Capstone Projects — Two larger end-to-end projects (the second is optional but counts toward the certificate).

Projects & Certificate

The course provides three project opportunities:

1. Midterm project (after Module 6)
2. Capstone project 1 (after Module 10)
3. Capstone project 2 (optional, at the end)

To qualify for a certificate during a live cohort, you must:

1. Submit two projects — either midterm + one capstone, or both capstones
2. Submit them by the cohort deadlines
3. Complete peer review of three other learners' projects

Past learner projects include a blood cell classifier for cancer prediction and a waste classifier (an Xception-based image classifier with 93.3% test accuracy, served through a Docker-packaged API).

Community

Learners connect through the DataTalks.Club Slack (#course-ml-zoomcamp channel), Telegram, and a YouTube playlist. The course encourages "learning in public" — sharing progress via blog posts, videos, and social media with #mlzoomcamp, which can earn bonus points during live cohorts.

Practical Notes

• The course is completely free and open source
• There's no powerful-computer requirement — cloud resources handle the heavy deep learning computation
• Self-paced learners can use all materials free, but certificates only come through live cohorts
• Real projects from the course give you concrete portfolio work for GitHub