mlops-get-started

Code snippets for LearnMLOps guides — practical MLOps examples for DevOps engineers.

Structure

mlops-get-started/
├── datasets/HR-Employee-Attrition.csv  # Dataset used across pipeline notebooks
├── 01-foundation/
│   ├── devops-to-mlops/           # 90-day learning plan
│   └── ml-basics/                 # First model notebook
├── 02-pipeline/
│   ├── dataset-pipeline/          # Ingestion, Pandera validation, Airflow DAGs, DVC
│   ├── data-preparation/          # Feature engineering, preprocessing, splits
│   └── ml-model-training/         # Training, MLflow tracking, Optuna, Kubernetes Jobs
├── 03-kubernetes/
│   ├── kubernetes-for-ml/         # Node affinity, taints, PVCs, Jobs, resource quotas
│   ├── kubernetes-gpu-workloads/  # NVIDIA device plugin, GPU Operator, MIG, DCGM
│   └── deploying-with-kserve/     # InferenceService, canary, autoscaling, monitoring
└── 04-operations/
    └── data-drift-model-decay/    # Drift detection, Evidently AI, DVC versioning, retraining

Dataset

The pipeline notebooks use the IBM HR Analytics Employee Attrition & Performance dataset from Kaggle.

Setup

→ SETUP.md — use the online sandbox or run locally with Docker Compose

Tool	Online sandbox	Local
Jupyter	https://notebook.ops4life.com	http://localhost:8888
MLflow	https://mlflow.ops4life.com	http://localhost:5000
Airflow	https://airflow.ops4life.com	http://localhost:8080

Usage

Clone and explore:

git clone https://github.com/ops4life/mlops-get-started.git
cd mlops-get-started

Each notebook corresponds to a code block in the LearnMLOps guides at learnmlops.ops4life.com. The pipeline notebooks (02-pipeline/) are runnable end-to-end in Google Colab — each includes !pip install and dataset fetch cells.