MLOps: Fundamentals of CI/CD and Model Deployment

Course details

As machine learning becomes central to modern software systems, DevOps engineers need new skills to manage models in production. This course introduces machine learning operations (MLOps) and covers how it extends DevOps practices to include data science workflows. Get an overview of the MLOps lifecycle—from CI/CD and continuous training to monitoring and governance—to learn how DataOps, ModelOps, and DevOps work together.

Find out how to collect and prepare data using tools like Pandas, Apache Spark, and Apache Kafka, then explore feature stores and pipeline orchestration with Airflow and Prefect. You get hands-on with MLflow for experiment tracking and model management, and BentoML for model deployment and serving. Along the way, learn about monitoring with Prometheus, Grafana, and Evidently, and how to address common data privacy, security, and compliance issues with GDPR, HIPAA, and PCI standards.

Concepts

Introduction to MLOps

Course introduction
Getting started with the machine learning team
Introduction to the MLOps engineer role
Comparison with DevOps - MLOps DataOps + DevOps + ModelOps
MLOps lifecycle
Continuous integration (CI), continuous deployment (CD), continuous training (CT), and continuous monitoring (CM)
Finding and exploring the right tools from DevOps for MLOps
MLOps architecture

Data Collection and Preparation

Data collection and preparation
Data ingestion - ETL
Data lakes
Data cleaning and data transformation
Small- and medium-sized datasets and data transformation - pandas and Polars
Demo - Small- and medium-sized datasets and data transformation pandas and Polars
Large datasets - Apache Spark (PySpark) and Dask
Streaming datasets - Apache Kafka and Apache Flink
Demo - Stream data using Apache Kafka
What is feature store
Data pipeline orchestration - Airflow and Prefect
Demo - Data pipeline orchestration

Model Development and Training

Model development
Model training and hyperparameter tuning
World of CPUs and GPUs
Introduction MLflow
Demo - Setting MLflow
Demo - Running an experiment and storing the result on MLflow
Demo - MLflow model artifact and versioning

Model Deployment and Serving

Model serving
Model drift and online offline serving
Model deployment and serving
Demo - Model serving using BentoML
Demo - Upgrading model versions with BentoML serving
Monitoring tools - Prometheus, Grafana, and Evidently

Automating Insurance Claim Reviews with MLflow and BentoML

Deploy app for insurance agents to upload all insurance claims
Demo - Generate dummy data for the project
Demo - Set up MLflow server and run the ML experiment
Demo - Register the model and setup BentoML for serving ML models
Demo - Upgrade Python Flask app to connect to BentoML for online serving

Data Security and Governance

Data privacy and data security
Data access management
Data retention
Need of compliance and GDPR
HIPAA compliance
PCI compliance
Compliance consequences and penalties
Compliance summary

Sneak Peek into AWS SageMaker

Overview of SageMaker
Core components of SageMaker
MLOps with SageMaker