Scalable Serverless Architecture for Delivering Personalized Recommendations

Recently, serverless computing has transformed the mode of development for scalable, cost-effective solutions for a great many applications. We are suggesting a scalable architecture for serving personalized recommendations that is based on native cloud services and microservices. This architecture dynamically allocates resources according to user demands in a scenario of lower operating costs with still high performance. Personalized recommendation systems have been of great importance to businesses, particularly in e-commerce, content streaming, and social platforms. Of course, the complexity lies in making all these recommendations available in real-time to millions of users without overloading the system. The serverless approach of this architecture decouples those core processing components, allowing individual microservices to scale independently based on traffic and computational demand. The system uses event-driven triggers for processing data and model inference as well as to generate a response, allowing for flexibility in a powerful way to grow workloads. We evaluate this architecture on several use cases, concentrated on the ones for e-commerce and content recommendation. Using both collaborative filtering and content-based approaches, we were able to achieve considerable improvements in terms of system scalability and low response times while reducing operational overheads.