MLOps Course
Take machine learning models from notebook to production: pipelines, deployment, monitoring, and retraining.
24 lessons · about 6 hours of reading · Free
Start lesson 1Course syllabus24 Lessons
What Is MLOps? From Notebooks to Reliable Production
Discover what MLOps is, why models fail outside notebooks, and how the end-to-end machine learning lifecycle bridges data science and software engineering.
MLOps Principles and Maturity: From Manual to Automated
Learn the six core principles of MLOps, evaluate systems using the standard three-level maturity model, and prioritize what to automate first.
ML Fundamentals for MLOps: Splits, Leakage, and Metrics
Master core ML concepts needed for MLOps: supervised learning pipelines, data splits, preventing leakage, evaluation metrics, and overfitting.
A Reproducible ML Project: Layout, Seeds, and Data Code
Learn how to structure a production ML repository, pin dependencies, manage configuration, and build a deterministic synthetic dataset generator in Python.
Data Engineering Basics for ML: Ingestion, SQL, and Skew
Understand where ML data originates, compare ETL and ELT patterns, write robust feature SQL, and prevent training-serving skew in production.
Pipeline Orchestration: Apache Airflow and Kubeflow
Learn how to schedule, retry, and manage multi-step ML pipelines reliably using DAGs, Apache Airflow, Kubeflow Pipelines, and cron.
Batch vs Streaming: Spark, Kafka, and Flink
Explore batch and streaming data architectures for machine learning. Understand Spark, Kafka, and Flink mechanics, and learn why simplicity wins before scale.
Data Versioning and Lineage: DVC and Beyond
Learn how to version datasets without bloating Git. Understand DVC content hashing, data lineage graphs, and the provenance triangle tying code to data and models.
Data Validation and Quality in ML Pipelines
Stop bad data from corrupting production models. Learn how to design data contracts, place quality gates, and build a tested validation engine.
Experiment Tracking with MLflow
Stop tracking ML experiments in spreadsheets and console logs. Learn how MLflow logs parameters, metrics, and artifacts, and build a tested run tracker.
Reproducible Training Pipelines
Turn exploratory notebooks into robust, deterministic ML training pipelines. Learn config-driven runs, atomic transformers, and model packaging.
Model Evaluation and Validation Gates
Prevent broken models from reaching production. Learn how to compare champion and challenger models, evaluate sub-population slices, and build automated gates.
Model Registry and Versioning
Manage machine learning models across their lifecycle. Learn registry concepts, lifecycle aliases, promotion gates, instant rollback, and build a testable registry.
CI/CD for Machine Learning
Automate testing, continuous training, and delivery for ML systems. Learn GitHub Actions workflows, CML pull request reports, and trigger architectures.
Containers and Kubernetes for ML
Package models into production containers and scale inference on Kubernetes. Learn Docker optimization, GPU pass-through, resource limits, and HPA.
Infrastructure as Code: Terraform and Ansible
Provision and configure repeatable ML cloud infrastructure. Learn Terraform state locking, cost guards, Ansible configuration, and environment guardrails.
Cloud ML Platforms: AWS, Azure, GCP, and Self-Built
Compare AWS SageMaker, Azure ML, and Google Cloud Vertex AI by capability, assess build vs buy trade-offs, and establish a platform decision framework.
Model Serving Patterns: Batch, Online, and Streaming
Master batch scoring, online REST/gRPC APIs, and streaming inference. Implement input validation, canary releases, and skew prevention.
Inference Engineering Basics: Latency, Batching, and Cost
Optimize machine learning inference systems for sub-second latency and minimal cloud spend. Master latency budgets, dynamic batching, and cost per prediction.
Monitoring and Observability for ML Models
Build end-to-end MLOps observability. Track system health, data integrity, and model quality using Prometheus, Grafana, and delayed ground truth.
Drift Detection and Retraining for ML Models
Detect data drift, concept drift, and covariate shift using PSI and statistical tests. Configure automated retraining pipelines and safe validation gates.
Explainable AI: LIME, SHAP, and Model Interpretability
Explain individual machine learning predictions using LIME and SHAP. Master global vs local explanations, feature attributions, and fairness audits.
Edge AI Deployment: Jetson, TFLite, and Mobile ML
Deploy machine learning models to edge hardware. Master TensorFlow Lite, PyTorch Mobile, NVIDIA Jetson, over-the-air updates, and fleet monitoring.
Capstone: An End-to-End MLOps Pipeline
Tie together the complete MLOps lifecycle for StayWell: data validation, training, gates, registry, serving, monitoring, and automated retraining.
Related courses
Get MLOps by email
One lesson a week, in order. Free, and you can unsubscribe from any email.
Frequently asked questions
Who is the MLOps course for?
Developers, data scientists and DevOps engineers who must get machine learning models into production and keep them there.
What do I need before starting the MLOps course?
Comfort with Python and Git. You do not need MLOps experience, and this is not a first programming course.
What will I learn in the MLOps course?
You will build an end-to-end MLOps pipeline covering training, deployment, monitoring and retraining.
How long does the MLOps course take?
It has 24 lessons and about 6 hours of reading, before you spend time on the practice tasks. Go at your own pace; the lessons are written to be done in order.
Is the MLOps course free?
Yes. Every lesson is free to read, and you do not need to sign in.