MLOps-Sentinel: Predictive Service & Monitoring Pipeline

Author: Adam Silva GitHub: https://github.com/adaumsilva/
Specialization: AI Engineering, RAG Architectures, and MLOps.

---

MLOps-Sentinel is a production-grade machine learning repository designed to demonstrate the complete lifecycle of a predictive model. Moving beyond experimental notebooks, this project implements a robust engineering framework for training, deploying, and monitoring high-availability models in a containerized environment.

---

Key Features

• Automated Training Pipeline: A modularized Python engine for data preprocessing, feature engineering, and Scikit-learn/XGBoost model training.
• Production API Layer: An asynchronous FastAPI implementation providing a /predict endpoint with strict Pydantic data validation.
• Containerization: Optimized multi-stage Docker builds to ensure consistent environments and small image footprints.
• CI/CD Integration: Automated GitHub Actions workflows for code linting (Black/Flake8) and unit testing on every push.
• Observability: Integration with Prometheus metrics to track inference latency and request volume in real-time.

---

System Architecture

The project follows a modular MLOps structure:

1. Data Ingestion: Raw data is processed through versioned transformation scripts.
2. Model Training: Models are trained and serialized using standard persistence formats.
3. Deployment: The API serves the model via a high-performance Uvicorn server.
4. Monitoring: Metrics are exposed via a /metrics endpoint for Prometheus scraping.

---

Tech Stack

• Model Framework: Scikit-learn / XGBoost
• API Framework: FastAPI / Uvicorn
• Validation: Pydantic
• Infrastructure: Docker / Docker-compose
• CI/CD: GitHub Actions
• Monitoring: Prometheus / Client_python
• Quality Assurance: Pytest

---

Getting Started

Prerequisites

• Python 3.10 or higher
• Docker and Docker Compose

Installation

1. Clone the repository:

   git clone [https://github.com/adaumsilva/MLOps-Sentinel.git](https://github.com/adaumsilva/MLOps-Sentinel.git)
   cd MLOps-Sentinel

2. Set up the environment:

   python -m venv venv
   source venv/bin/activate  # Windows: venv\Scripts\activate
   pip install -r requirements.txt

3. Train the initial model:

   python src/training/train.py

Running with Docker

To deploy the API and Prometheus monitoring stack simultaneously:

docker-compose up --build