Automated Machine Learning Deployment System Using MLOps

Abstract

The increasing adoption of Machine Learning (ML) in real-world applications has exposed significant challenges in deploying models reliably and reproducibly at scale. While most ML workflows primarily focus on model development and evaluation, operational aspects such as experiment tracking, version control, automated retraining, and continuous deployment are often inadequately addressed. This paper presents an Automated Machine Learning Deployment framework grounded in MLOps principles to bridge the gap between model development and production deployment. The proposed system automates the end-to-end ML lifecycle, including data ingestion, multi-model training, performance evaluation, and best-model selection. Data Version Control (DVC) is employed to manage dataset versions, while batch-based data drift monitoring is implemented using Evidently AI to analyze distributional changes across training iterations. MLflow is utilized for systematic experiment tracking and model versioning, ensuring reproducibility and traceability of results. Continuous Integration and Continuous Deployment (CI/CD) are implemented using GitHub Actions to enable automated retraining and seamless deployment upon code or data updates. The selected model is containerized using Docker and deployed as a FastAPI-based REST service on an AWS EC2 instance. The proposed framework demonstrates improved deployment reliability, scalability, and reproducibility through automated MLOps practices, making it suitable for real-world machine learning applications.