Exclusive: In the race for robot training data, ex-Meta researchers’ startup turns to gig workers with cameras

Memories.ai, a San Francisco-based startup, is paying gig workers to record first-person video footage of daily tasks like food preparation, shelf stocking, and equipment repair. The company's wearable LUCI Pin camera, weighing under 45 grams, captures POV video of contributors performing various tasks while attaching magnetically to mounts on a person's head, chest, or wrist.

The platform, Memories Upload, aims to collect over 10 million hours of video data by March 2027, having already recorded more than 1 million hours in its first month of beta. Founded by former Meta researchers Shawn Shen and Ben Zhou, Memories.ai has raised $16 million in venture funding, including an $8 million seed round led by Susa Ventures and an additional $8 million in June 2026.

The company collaborates with humanoid robotics startup Figure and Physical Intelligence, which builds AI software for controlling various robots and embodied AI devices. By partnering with Nvidia, Memories.ai utilizes Metropolis Video Search and Summarization tools and fine-tuned video-captioning models to optimize video processing.

Co-founder and CEO Shawn Shen emphasizes the importance of diverse, camera-angle-rich footage for robot learning, stating that even seemingly similar tasks like making coffee can provide valuable variation when captured from different environments and perspectives. Privacy concerns are addressed by requiring consent forms from camera wearers and coworkers who regularly appear in recordings, with automated and manual processes in place to obscure sensitive details.