Home

🧠 MLOps Platform Wiki

Welcome to the MLOps Platform wiki – your central guide to understanding, deploying, and extending a production‑grade machine learning infrastructure.

---

📖 What is this platform?

This repository implements a full MLOps stack that automates the entire machine learning lifecycle:

• Feature Engineering using Feast (offline & online store)
• Experiment Tracking & Model Registry with MLflow
• Training Pipeline Orchestration via Kubeflow Pipelines
• Advanced Drift Detection (KS‑test, Jensen‑Shannon, PCA) for automatic retraining
• Model Serving with KServe and GPU‑optimised transformers
• Infrastructure as Code (Terraform) on AWS EKS
• CI/CD with GitHub Actions

It is designed to be cloud‑agnostic, scalable, and cost‑efficient, supporting teams that need to move from experimentation to production with confidence.

---

🗺️ Architecture at a Glance

graph TD
    subgraph "Data"
        A[Data Lake / S3]
        B[Streaming Events]
        C[Feast Feature Store]
    end

    subgraph "ML Pipelines (Kubeflow)"
        D[Training Pipeline]
        E[Drift Detection]
    end

    subgraph "MLflow"
        F[Experiment Tracking]
        G[Model Registry]
    end

    subgraph "Serving (KServe)"
        H[InferenceService]
        I[Transformer]
        J[Predictor]
    end

    subgraph "Infrastructure (EKS)"
        K[Kubernetes]
        L[GPU Nodes]
        M[Prometheus Monitoring]
    end

    A --> C
    B --> C
    C --> D
    D --> F
    F --> G
    G --> H
    H --> I --> J
    E --> D
    K --> L
    M --> K

Loading

For a detailed walk‑through see the Architecture Deep-Dive page.

---

🚀 Quick Start

1. Clone the repository

   git clone https://github.com/Awrsha/mlops-platform.git
   cd mlops-platform

2. Provision infrastructure with Terraform

   cd terraform
   terraform init && terraform apply

3. Deploy cluster services (KServe, MLflow, Feast, Kubeflow)

   kubectl apply -f kubernetes/

4. Trigger a training pipeline
   Submit the compiled pipeline to your Kubeflow endpoint or let the CI/CD handle it automatically.

For a more detailed setup, read the Installation Guide.

---

📚 Wiki Pages

• Installation Guide – Step‑by‑step instructions to deploy the platform.
• Architecture Deep-Dive – Component design, data flow, and scaling strategies.
• Drift Detection – How statistical tests trigger automatic retraining.
• CI/CD Pipelines – GitHub Actions workflows explained.
• Model Serving – KServe configuration, transformers, and GPU optimisation.
• Monitoring & Alerting – Prometheus rules, Grafana dashboards, and alerts.
• Troubleshooting – Common issues and solutions.

---

🤝 Contributing

We welcome contributions! See the main README or the Contributing Guide for details.

---

❓ Need Help?

Open a GitHub Issue or reach out to the team on Discussions.

Happy building! 🚀