Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Discover how a small, specialized AI model revolutionizes video conferencing technology in this 47-minute conference talk from NDC AI. Learn the compelling story of how a frustrated engineer's quest to improve existing software led to the development of a tiny head detection model that now powers an entire intelligent video conferencing platform. Explore the counterintuitive principle that in a world obsessed with massive language models and billion-parameter systems, sometimes the most impactful AI solutions are surprisingly small and focused. Understand how knowing exactly where people are positioned and which direction they're facing becomes the foundational information that enables speaker tracking, automatic framing, people counting, and camera control features. Examine the technical considerations and constraints involved when deploying AI models on edge devices in compute- and time-limited environments. Gain insights into how a specialized, efficient model can serve as the backbone for an entire suite of intelligent video conferencing features, demonstrating that targeted solutions often outperform complex alternatives in real-world applications.