**Sieve is an AI compute platform for video. Our goal is to build useful primitives to understand, manipulate, and create video so we can accelerate all applications of it.**
Today, video makes up over 82% of internet traffic and every application of it is growing extremely quickly — media, communications, VR, gaming, robotics, automotive, etc. TikTok went from 0 → 2B users in the last decade. Video conferencing is used many times a day by every knowledge worker. Google’s Veo 2 model generates physically realistic video. Waymo now has more market share than Lyft in SF ride share. Tesla is scaling humanoid robots.
As these applications grow, the need to process video grows too. But unlike other data types, video is complex, and applications of it require purpose-built pipelines to tackle specific problems. Some examples:
At Sieve, we engineer production-grade pipelines to understand, manipulate, and create video — along with flexible abstractions to run custom pipelines. These pipelines already power core features in some of the most popular creative and social video platforms in the world, and are also helping research teams collect & annotate large datasets to train better video models.
The unique complexity of video requires both R&D and infrastructure that is intricate, expensive, and time-consuming to build in-house. Sieve comes with a set of production-grade pipelines engineering teams don’t need to worry about building or maintaining, along with infrastructure abstractions that make it easy to develop increasingly complex and specific video processing tasks on top.
Everyone at Sieve cares deeply about the work they do. You will enjoy working here if you: