Recommender systems are an integral part of most products nowadays and are often a key driver of discovery for users of the product. Developing a large-scale recommender system that can provide personalized recommendations to several hundred million users while ensuring relevance and timeliness is as much of a machine-learning problem as it is an engineering task. Additionally, with the recent advancement in the AI space, we are able to train larger and larger models that can capture users' long-term preferences, which ultimately improves the personalization of such recommender systems. However, building and integrating such large foundation models into the recommender systems is yet another machine learning and engineering task that requires a lot of optimization of the algorithms, as well as the training and inference process.

In this talk, I will cover building such Recommender Systems and Foundation Models that can power products for multi-million users as well as highlight challenges faced, things to be considered, and success stories in real-world applications. This talk will bring the engineering and infrastructure perspective on these tasks, while also highlighting the AI and machine learning perspective.