Amagi built a real-time AI pipeline that watches live news streams and analyses audio and video simultaneously, capturing shareable storylines as they unfold, with audio processed in under 200 milliseconds and video in under two seconds. Mayank Sharma, Principal Architect at Amagi, explained in an Amazon Web Services (AWS) GenAI Tech Talks episode that the team used Amazon SageMaker to handle infrastructure, scaling, and security, reducing costs by roughly 50% compared to always-on GPU infrastructure. The system scales automatically across more than 50 streams during breaking news and down to five during off-peak hours without manual intervention.