Fast Christmas: Production-Grade ML Inference Service

Fast Christmas is a high-performance, containerized microservice built to deliver real-time car price predictions through a robust FastAPI backend. The project demonstrates a full-lifecycle engineering approach, from an automated machine learning pipeline to secure, observable deployment in a microservices environment.

Engineering Highlights

Automated ML Pipeline & Persistence: Developed a structured training workflow using Scikit-Learn pipelines, incorporating ColumnTransformer for automated numerical scaling and categorical encoding. The final Random Forest Regressor is persisted via joblib for seamless integration into the inference service.

High-Performance Caching:

Optimized API latency by implementing a Redis caching layer. The service intelligently caches prediction results based on input feature hashes, significantly reducing redundant compute cycles for recurring queries.

Enterprise-Grade Security:

Architected a dual-layer security model utilizing JWT (JSON Web Tokens) for user session management and custom Header-based API Key validation for secure service-to-service communication.

Full-Stack Observability:

Integrated a comprehensive monitoring suite using Prometheus and Grafana. The service exports real-time performance metrics via prometheus-fastapi-instrumentator and includes custom logging middleware to track request/response lifecycles.

Microservices Architecture:

Fully containerized using Docker and Docker Compose, orchestrating the API, Redis, Prometheus, and Grafana into a cohesive, scalable backend.

Resilient Design:

Implemented global exception handling and structured Pydantic data validation to ensure API stability and consistent error reporting (500-series handling).

Tech Stack

Backend:

Python 3.12, FastAPI, Pydantic, Uvicorn.

Machine Learning:

Scikit-Learn (Random Forest), Pandas, Joblib.

Data & Infrastructure:

Redis (Caching), Docker, Docker Compose.

DevOps & Monitoring:

Prometheus, Grafana, Starlette Middleware.

Security:

Python-Jose (JWT), Dotenv.

---

Getting Started (Local)

1. Clone the Repository

git clone https://github.com/your-username/fastapi-project.git
cd fastapi-project

2. Set Environment Variables

API_KEY=demo-key
JWT_SECRET_KEY=your-secret
REDIS_URL=redis://localhost:6379

3. Build and Run via Docker

docker-compose up --build

4. Access Interfaces

• FastAPI Docs: http://localhost:8000/docs
• FastAPI Metrics: http://localhost:8000/metrics
• Prometheus UI: http://localhost:9090
• Grafana UI: http://localhost:3000

---