MLOps Learning Guide¶

Overview¶

MLOps (Machine Learning Operations) is a practical field that automates and streamlines the development, deployment, and operation of machine learning models. This learning material is designed for learners who have completed the basics of Deep Learning and covers the overall processes and tools needed to operate ML systems reliably in production.

Target Audience¶

Those who have completed Deep Learning basics
Developers who want to deploy ML models to production
Software engineers collaborating with data science teams
DevOps/SRE engineers responsible for operating ML systems

Learning Roadmap¶

MLOps Overview → ML Lifecycle → MLflow Basics → MLflow Advanced → W&B
     │                              │              │         │
     │                              ↓              ↓         ↓
     │                         Experiment Tracking ──── Model Registry ──→ Kubeflow
     │                                                            │
     ↓                                                            ↓
Feature Store ← Drift/Monitoring ← TorchServe/Triton ← Model Serving Basics
     │
     ↓
Practical MLOps Project (E2E Pipeline)

File List¶

File	Topic	Difficulty	Key Content
01_MLOps_Overview.md	MLOps Overview	⭐	MLOps definition, DevOps vs MLOps, maturity levels, tool ecosystem
02_ML_Lifecycle.md	ML Lifecycle	⭐⭐	Project phases, retraining triggers, version control strategies
03_MLflow_Basics.md	MLflow Basics	⭐⭐	Tracking, experiment management, metrics/parameters logging, UI
04_MLflow_Advanced.md	MLflow Advanced	⭐⭐⭐	Projects, Models, Registry, Serving
05_Weights_and_Biases.md	Weights & Biases	⭐⭐	Experiment logging, Sweeps, Artifacts, MLflow comparison
06_Kubeflow_Pipelines.md	Kubeflow Pipelines	⭐⭐⭐	Pipeline SDK, component authoring, K8s integration
07_Model_Registry.md	Model Registry	⭐⭐	Version control, stage transitions, CI/CD integration
08_Model_Serving_Basics.md	Model Serving Basics	⭐⭐	REST API, gRPC, batch vs real-time inference
09_TorchServe_Triton.md	TorchServe & Triton	⭐⭐⭐	Handler authoring, model optimization, multi-model serving
10_Drift_Detection_Monitoring.md	Drift & Monitoring	⭐⭐⭐	Data/model drift, Evidently AI, alert configuration
11_Feature_Stores.md	Feature Store	⭐⭐⭐	Feast, online/offline stores, feature serving
12_Practical_MLOps_Project.md	Practical MLOps Project	⭐⭐⭐⭐	E2E pipeline, automated retraining, project structure

Environment Setup¶

Installing Required Libraries¶

# Basic ML libraries
pip install numpy pandas scikit-learn torch torchvision

# MLOps tools
pip install mlflow wandb feast evidently

# Serving tools
pip install torchserve torch-model-archiver

# Kubeflow (Python SDK)
pip install kfp

Docker Setup¶

# MLflow server (Docker)
docker run -d \
  --name mlflow-server \
  -p 5000:5000 \
  -v $(pwd)/mlruns:/mlruns \
  ghcr.io/mlflow/mlflow:latest \
  mlflow server --host 0.0.0.0 --backend-store-uri /mlruns

# Triton Inference Server
docker run --gpus all -d \
  --name triton-server \
  -p 8000:8000 -p 8001:8001 -p 8002:8002 \
  -v $(pwd)/model_repository:/models \
  nvcr.io/nvidia/tritonserver:23.10-py3 \
  tritonserver --model-repository=/models

Version Check¶

import mlflow
import wandb
import feast

print(f"MLflow: {mlflow.__version__}")
print(f"W&B: {wandb.__version__}")
print(f"Feast: {feast.__version__}")

Recommended Versions¶

Python: 3.9+
MLflow: 2.8+
Weights & Biases: 0.16+
Feast: 0.35+
Kubernetes: 1.25+

Recommended Learning Order¶

Stage 1: Basic Theory (01-02)¶

Understand MLOps concepts and necessity
Learn ML project lifecycle

Stage 2: Experiment Management (03-05)¶

Track experiments with MLflow
Utilize Weights & Biases
Manage metrics, parameters, and artifacts

Stage 3: Pipelines (06-07)¶

Build ML pipelines using Kubeflow
Operate model registry

Stage 4: Model Serving (08-09)¶

Deploy REST/gRPC APIs
Utilize TorchServe, Triton

Stage 5: Monitoring & Feature Store (10-11)¶

Detect drift
Build feature stores

Stage 6: Practical Project (12)¶

Build E2E MLOps pipeline
Implement automated retraining system

Official Documentation¶

Recommended Books¶

"Introducing MLOps" - Mark Treveil
"Machine Learning Engineering" - Andriy Burkov
"Designing Machine Learning Systems" - Chip Huyen

Docker folder: Container basics
Deep_Learning folder: Deep learning model training
Machine_Learning folder: ML algorithm basics