MLOps-ResearchProject-TFM

An end-to-end MLOps platform for dynamic MLflow model deployment, monitoring, and management. This project upgrades the TFG to MLOps Level 2, providing a production-ready, containerized solution for model lifecycle automation.

🚀 Overview

This system enables: - Dynamic deployment of MLflow models with automatic port allocation - Centralized control panel (React frontend) for model management - Metrics tracking, monitoring, and reporting - API-first architecture for integration and automation

🏗️ Architecture

Main components: - Frontend: React dashboard for model management and monitoring - Model Deployment Service: FastAPI backend for deploying, undeploying, and proxying models - MLflow Tracking Server: Model registry and artifact store - Databases: PostgreSQL (MLflow), MongoDB (metrics/data) - Nginx: API gateway, authentication, and routing

User → Nginx → Model Deployment API → MLflow/Deployed Models

✨ Features

• Deploy/undeploy MLflow models on demand
• Automatic port management (8001-8100)
• Model versioning and registry integration
• Real-time metrics and health monitoring
• Download datasets and compare model metrics
• Secure access via nginx basic auth
• Responsive, modern UI

⚡ Quick Start

Prerequisites

• Docker
• Docker Compose
• AWS account for S3 recomended, GCP account for HDFS or having a local instance of HDFS.
• Two telegram bots.

Installation

Clone the repo and run:

bash ./install.sh

Once installed you could run with:

bash docker compose up -d

The system will be available at http://localhost (see nginx config for details).

🖥️ Usage

Web UI

1. Go to http://localhost (or your configured domain)
2. Log in with your credentials (nginx basic auth)
3. Use the dashboard to:
   • View available models
   • Deploy/undeploy models
   • Monitor deployed models and metrics
   • Download datasets

API Endpoints (examples)

• List models: GET /get_model_list
• Deploy model: POST /deploy/{model}/{version}
• List deployed models: GET /get_deployed_models
• Undeploy model: POST /undeploy/{model-version}
• Call model: POST /{model}-{version}

See MODEL_DEPLOYMENT_README.md for full API docs and usage examples.

📁 Directory Structure

• containers/frontend/ – React app (dashboard)
• containers/model_deployment/ – FastAPI backend for deployment
• containers/mlflow/ – MLflow tracking server
• containers/mongodb/ – MongoDB init scripts
• nginx/ – Nginx config for routing/auth
• examples/ – Example training scripts and notebooks
• compose.yaml – Docker Compose setup

🛠️ Development

Frontend

Backend

NGINX

mlflow

MongoDB

Postgres

Remote logs

🩹 Troubleshooting

• Model not accessible: Verify deployment, check logs, ensure correct model name-version
• Frontend/API errors: Check container logs, nginx config, and authentication

🤝 Contributing

Pull requests and issues are welcome! Please open an issue to discuss major changes.

📄 License

MIT License

📬 Contact

For questions or support, open an issue or contact the maintainer.