My Next Big Challenge

Ten years ago, there was very little in the way of tools and infrastructure for high-quality video encoding at scale, and everything we did was a first. In times of solving big problems we often think little about the efficiency of our tools, [rightly] focusing on features and functionality. For video, this has resulted in a very large compute footprint, and often wasteful use of compute capacity, and the problem is getting worse. Video consumption is growing significantly. Further, convex hull encoding adds significant computational complexity to video encodes, and each new generation of video encoders carries a 6-10x increase in computational complexity. At the same time, the need for compute-heavy tools like video understanding (computer vision) also require significant compute capacity. I believe that one of the biggest video industry challenges of the next decade is reducing the compute and storage requirements for video encoding; bending the compute growth curve down, while still maximizing quality and codec efficiency. Reducing the cost of encoding is good business. However, we also have a responsibility to minimize the environmental costs of our work, and thus we must put priority on reducing wasteful use of compute capacity, and finding low-power solutions to many video processing tasks.  

I recently joined Facebook as a leader in the Video Infrastructure org. Having completed Bootcamp, it’s time to have fun! Reducing the compute and storage requirements for high-quality video encoding is just one of the many exciting projects I will be working on.  It's a huge challenge and I am very excited to be on the Facebook team.