Memories.ai is building the visual memory layer for wearables and robotics

Memories.ai, founded by ex-Meta engineers who built AI for RayBan glasses, is developing a large visual memory model (LVMM) that enables wearables and robots to store and recall video-based memories. The startup announced a collaboration with Nvidia at GTC 2026, leveraging Cosmos Reason 2 and Nvidia Metropolis for video search and summarization. Unlike text-based memory systems from OpenAI, Google, and xAI, Memories.ai focuses on visual memory for physical AI applications. The company has raised $16M, launched its LVMM in July 2025, built a custom data-collection wearable called LUCI, and partnered with Qualcomm to run its model on Qualcomm processors. It is already working with unnamed large wearable companies while positioning itself primarily as a model and infrastructure provider.