Why it matters
A mix of applied talks from a robotics PhD student, TwelveLabs, Voxel51 and Nebius. The video retrieval and embedding compression talks cover problems that matter for anyone running retrieval at scale.
The tokenmaxxing angle
The link is loose. One talk covers compressing LLM embeddings, which shrinks vector storage and retrieval cost. Another covers video-language model search, where per-hour processing cost is the question to bring.
From the organizers
Co-presented by Nebius and Voxel51 at Fractal Tech, 111 Conselyea St. Talks include Irmak Guzey (NYU) on dexterous robot hands, Yash Gugale (TwelveLabs) on video search, and Dhruv Diddi (Nebius) on physical AI.