A full-stack MLOps platform that lets anyone regardless of ML experience train, compare, deploy, and run inference on machine learning models through a guided step-by-step web interface. Built on Django, ZenML, MLflow, and AWS.
Can be run locally without AWS using local file storage for models and artifacts.
🎥 
- Upload a CSV and train multiple ML models through a step-by-step UI
- Experiment tracking and model registry via MLflow
- Deploy models as persistent HTTPS inference endpoints
- Run batch predictions and store results
- Session-based Guest accounts. Registered users retain full history
- CI pipeline via GitHub Actions to AWS ECR
The app runs as a multi-container system via Docker Compose on an AWS EC2 instance:
| Service | Technology | Purpose |
|---|---|---|
| Web Server | Django + Gunicorn | Frontend, auth, API views |
| ML Tracking | MLflow | Experiment logging, model registry |
| ML Pipelines | ZenML | Pipeline orchestration |
| App Database | PostgreSQL | User data, models, files |
| ZenML Database | MySQL | ZenML metadata |
| File Storage | AWS S3 | CSVs, plots, static files |
| Reverse Proxy | nginx | HTTPS routing between containers |
- Backend: Django, Gunicorn
- ML: Scikit-learn, ZenML, MLflow
- Cloud: AWS (EC2, S3, ECR, SSM Parameter Store)
- Infrastructure: Docker, Docker Compose, nginx
- CI: GitHub Actions
- Databases: PostgreSQL, MySQL
- Languages: Python 3.13, Bash
- Docker and Docker Compose
- AWS account with ECR, S3, and SSM configured (OPTIONAL, CAN USE LOCAL FILE STORAGE)
- Clone the repository
git clone https://github.com/reyrey112/MLAPP.git
cd MLAPP- Add your environment variables
cp .env.example .env
# Fill in your values- Run the app
docker-compose up -d- Visit
http://localhost:8000
MLAPP/
├── MLapp/ # Django project settings
├── processing/ # Django app (views, models, templates)
├── MLOps/
│ ├── pipelines/ # ZenML pipeline definitions
│ └── steps/ # ZenML step definitions
├── .github/workflows/ # CI/CD GitHub Actions
├── docker-compose.yaml
├── django_dockerfile
├── nginx.conf.template
└── ec2_startup_script.sh
Why ZenML alongside MLflow? MLflow handles experiment tracking and model registry well, but ZenML provides pipeline orchestration, step-level caching, and artifact lineage that MLflow alone doesn't offer. Together they cover the full MLOps lifecycle from training to deployment.
As a pipeline orchestrator, ZenML allows integrations between many different softwares (local and cloud-based) and can create fully customizable software stacks for different pipeline executions that can be changed seamlessly, making it perfect for fitting into anybody's current tech stack.
- Automate EC2 startup via User Data and crontab
- A more robust Continuous deployment pipeline for new ECR images
- TensorFlow model support
- Expanded visualization library
MIT