MLOps for Improved Inferencing, Deployability and Observability of Recommendation Engine

One of the major goals of over the top (OTT) platform is to provide best user experience to its customer and recommendation system plays a major role to achieve the same. There are multiple machine learning models which are used to recommend available contents based on user history, click events etc. These models keep evolving over time due to changing dataset and continuous learning. MLOps process helps organization to manage, deploy and monitor these models in real production environment in an ease and effective way. This paper gives detail insight about MLOps process using open source MLflow framework and other supported tools, covering all stages from data ingestion to model deployment for a simple recommendation model to improve productivity and reduce development and operation cost.