International Journal of Scientific Research and Engineering Development

With the rise of digital media, audiences find the choices overwhelming, thereby making personalized recommendations a crucial aspect of streaming services today. OTT services, including Netflix, YouTube, Spotify, and Hotstar, have increasingly deployed recommendation systems to guide users toward content that they feel most attracted to. Going over how the systems have changed-from early days of pure collaborative and content-based filtering to hybrid and deep learning approaches-these systems represent widely used two-stages retrieval and ranking. The study inquires how some of the leading platforms fill in this architecture, solving their particular challenges; be it Netflix's foundation models, YouTube's large-scale neural networks, Spotify's hybrid ecosystem, or Hotstar's real- time, regionally aware personalization. It looks at key challenges, such as cold-start, data sparsity, algorithmic bias, privacy issues, and potential solutions such as diversity- aware ranking and federated learning. Finally, it attempts a look at potential futures- foundation with large language models, explainable AI, and causal inference-which are beginning to shape the recommendation system.