Interactive MLOps Course with Claude Code

Learn production machine learning by building and deploying a real sentiment classifier. This hands-on course uses Claude Code as your AI learning partner, guiding you through the complete MLOps lifecycle.

What You'll Build

A production-ready sentiment classifier for customer feedback analysis. You'll take it from a local script to a fully deployed, monitored API with CI/CD.

                                YOUR LEARNING JOURNEY

    Level 1               Level 2               Level 3               Level 4
   Foundations            Pipeline             Deployment           Production

   ┌─────────┐          ┌─────────┐          ┌─────────┐          ┌─────────┐
   │  Local  │          │   DVC   │          │ FastAPI │          │  CI/CD  │
   │  Model  │────────▶ │ + MLflow│────────▶ │   API   │────────▶ │ Pipeline│
   └─────────┘          └─────────┘          └─────────┘          └─────────┘
        │                    │                    │                    │
        ▼                    ▼                    ▼                    ▼
   ┌─────────┐          ┌─────────┐          ┌─────────┐          ┌─────────┐
   │ Docker  │          │  Tests  │          │  Cloud  │          │  Drift  │
   │ Package │          │         │          │ Deploy  │          │ Monitor │
   └─────────┘          └─────────┘          └─────────┘          └─────────┘

Prerequisites

• Python proficiency (comfortable with classes, modules, virtual environments)
• Basic ML concepts (sklearn, pandas, train/test splits)
• Command line familiarity
• Docker installed
• Claude Code CLI installed (installation guide)

Getting Started

1. Clone this repository:

   git clone https://github.com/sshtomar/ai-pair-programming-mlops.git
   cd ai-pair-programming-mlops

2. Start Claude Code:

   claude

3. Begin the first lesson:

   /start-1-1
   

Course Structure

Level 1: Foundations

Lesson	Topic	What You'll Learn
1-1	Welcome & Setup	Environment setup, course navigation
1-2	MLOps Principles	Why 87% of ML projects fail, the 5 pillars of MLOps
1-3	Your First Model	Build a sentiment classifier with production structure
1-4	Packaging for Production	Dockerize your ML application

Level 2: The ML Pipeline

Lesson	Topic	What You'll Learn
2-1	Data Versioning	DVC setup, linking data to code versions
2-2	Experiment Tracking	MLflow for parameters, metrics, artifacts
2-3	Model Registry	Model lifecycle, staging to production workflows
2-4	Testing ML Code	Unit tests, integration tests, behavioral testing

Level 3: Deployment

Lesson	Topic	What You'll Learn
3-1	Model Serving Options	Batch vs real-time, latency vs throughput
3-2	Building an API	FastAPI endpoints, validation, health checks
3-3	Containerization Deep Dive	Multi-stage builds, optimization, secrets
3-4	Deploying to Cloud	Cloud Run deployment, auto-scaling

Level 4: Production Operations

Lesson	Topic	What You'll Learn
4-1	CI/CD for ML	GitHub Actions, automated testing, deployment gates
4-2	Monitoring & Observability	System vs model metrics, logging, alerting
4-3	Model Drift & Retraining	Drift detection, retraining triggers
4-4	Capstone	End-to-end review, architecture assessment

Commands

Run these in Claude Code:

Command	Description
/start-X-Y	Begin lesson X.Y (e.g., /start-1-1)
/status	Check your course progress
/check	Verify your exercise solution
/hint	Get a hint for the current exercise
/review-code	Get ML engineer code review
/review-deployment	Get SRE deployment review
/help-mlops	Get help on MLOps concepts

How This Course Works

This course uses a Socratic teaching approach:

• Claude asks questions to check your understanding before explaining concepts
• You attempt exercises before receiving help
• Hints are provided in stages, not as complete solutions
• Real-world production concerns are emphasized throughout

The project/ directory contains your working codebase. Each lesson builds on previous work, so complete them in order.

Estimated Time

Approximately 12 hours for the full course, though you can complete lessons at your own pace.

License

MIT