Portfolio Details

Each project in this portfolio reflects my hands-on experience in solving real-world problems using data science, machine learning, and full-stack development. From predictive models to intelligent web applications, these solutions are built with a focus on scalability, functionality, and impact.

Project Information

  • Project Name: Perfume Haven
  • Category: MLOps Integration
  • Model Type: TF-IDF + Cosine Similarity
  • Github Repo: Link

Technologies Used

This project demonstrates full-scale MLOps practices using tools like MLflow for experiment tracking, DVC for data and pipeline versioning, and GitHub Actions for automated CI/CD workflows. The backend model is containerized using Docker and deployed via AWS EC2 and EKS (Kubernetes). Monitoring is handled with Prometheus and visualized in Grafana, ensuring performance reliability and observability in real-time.

MLOps Workflow Overview

Perfume Haven follows an end-to-end MLOps architecture to ensure consistency, automation, and scalability in production. Raw data is ingested and stored in AWS S3, where DVC pipelines handle preprocessing, feature extraction, and model training. Experiments are tracked and logged via MLflow, with the best-performing model automatically registered.

The trained model is served using FastAPI inside a Docker container. CI/CD workflows — managed through GitHub Actions — automate testing, Docker builds, and deployment to AWS EKS (Kubernetes). Monitoring is established through Prometheus for metrics collection and Grafana for real-time alerting and visualization.

MLOps Stack Highlights

  • MLflow for experiment tracking and model registry
  • DVC for data, code, and pipeline versioning
  • GitHub Actions for CI/CD automation
  • Docker containerization for model serving
  • AWS: S3 (storage), EC2 (compute), EKS (deployment)
  • Prometheus & Grafana for real-time monitoring & alerts

This project is a complete demonstration of real-world MLOps — from raw data to live deployment — integrating model reproducibility, automated delivery, and operational monitoring at scale.