Memories.ai
"AI that sees, remembers and acts"
US AI lab building a Large Visual Memory Model, a visual memory layer that lets AI, wearables and robots see, remember and recall video.
Memories.ai is a US AI research company that launched what it describes as the world's first Large Visual Memory Model (LVMM), giving machines long-term visual memory to index, search, summarize and recall millions of hours of video. The company was founded in 2024 by former Meta Reality Labs researchers Dr. Shawn Shen (CEO) and Ben Zhou (CTO), who conceived the idea while building the AI system behind Meta's Ray-Ban smart glasses before spinning out. The LVMM launched in July 2025 and a second generation followed. Memories.ai positions its visual memory infrastructure explicitly for physical AI, with a dedicated robotics and wearables vertical, and has partnered with Nvidia (using Cosmos-Reason 2 and Metropolis) and Qualcomm (on-device deployment on Snapdragon and X2 Elite NPU).
It built an in-house wearable, LUCI, for training-data collection rather than as a sold product, and demonstrated Project LUCI at Microsoft Build 2026 showing on-device memory across PC, AR glasses and home cameras. The company raised approximately 16 million dollars total, including an 8 million dollar seed in July 2025 led by Susa Ventures with participation from Seedcamp, Fusion Fund, Crane Venture Partners, Samsung Next and Creator Ventures. Its shipping commercial tools include video search, video understanding, summarization and agentic editing, but its stated mission targets memory infrastructure for embodied and physical AI, wearables and robots. The company is GDPR compliant and SOC 2 Type II certified.
Primary type & automation activities this supplier delivers:
Large Visual Memory Model
- Model: LVMM
- Launch Year: 2025
World's first large visual memory model; indexes and retrieves ultra-long video as recallable memory for AI and physical AI.
Project LUCI
- Model: LUCI
- Launch Year: 2026
Local-first on-device visual memory unifying personal memory across PC, AR glasses and home cameras.
Video Search AI
Natural-language search across large video archives to locate clips and incidents instantly.
Contact Memories.ai
WEBSITE
https://memories.aiHEADQUARTERS
United States
Company Facts
Founded
2024
Primary Role
Software/Algorithm
Company Size
-
Primary Region
North America
Annual Sales
-
Funding Stage
-
Funding Total
-
Certifications & Memberships
Death to Proprioception: How Vision-Only Robots are Conquering Spatial Generalization
A September 24, 2025 study by Spirit AI introduces the “State-free Policy,” eliminating proprioceptive inputs in visuomotor control to boost robotic spatial generalization. Using dual wide-angle wrist cameras and relative end-effector actions, success rates improved from near zero to 85% in real-world manipulation tasks.
