Memories.ai Builds Visual Memory Layer for AI Wearables at NVIDIA GTC 2026
Key Takeaways
- Memories.ai partners with Nvidia to build a visual memory layer for wearables and robotics.
- Developing a large visual memory model to index and retrieve video memories for AI systems.
- Announced at Nvidia GTC 2026, highlighting AI infrastructure and edge devices.
Startup Announcement
Memories.ai emerged as a new AI infrastructure player at NVIDIA GTC 2026.
“Shawn Shen believes that AI will need to remember what it sees in order to succeed in the physical world”
The startup introduced a large visual memory model for robots and wearables.
The company positions itself at the intersection of AI infrastructure and physical AI markets.
According to The Tech Buzz, Memories.ai tackles physical AI's memory challenge.
The startup's solution treats video recordings as a searchable database of experiences.
Leadership & Partnership
Shawn Shen believes AI needs visual memory to succeed in the physical world.
Memories.ai announced an NVIDIA collaboration at GTC 2026.
The partnership uses Nvidia's Cosmos-Reason 2 and Metropolis technologies.
Shen and Zhou developed the concept while building Meta's Ray-Ban glasses AI.
They couldn't find existing visual memory solutions, so they spun out from Meta.
Technical Challenge
Visual memory represents a significant technical challenge in AI development.
“Shawn Shen believes that AI will need to remember what it sees in order to succeed in the physical world”
Large language models excel at text, but visual memory indexing remains unsolved.
The Tech Buzz identifies this as the gap Memories.ai is targeting.
The timing is advantageous as the physical AI market expands rapidly.
Memories.ai's platform sits between camera sensors and AI applications.
Memory Landscape
Text-based memory systems have advanced significantly in recent years.
OpenAI, xAI, and Google Gemini have launched memory tools.
Shen notes these advancements focus on text-based memory.
Text memory is more structured but inadequate for physical AI needs.
Physical AI systems interact primarily through sight and visuals.
Applications & Market
Memories.ai targets specific physical AI applications.
“Shawn Shen believes that AI will need to remember what it sees in order to succeed in the physical world”
Use cases include AI glasses remembering key locations.
Warehouse robots can recall inventory layouts.
The infrastructure handles visual memory processing complexity.
The company positions itself at the forefront of a critical market bottleneck.
More on Technology and Science
Facebook launches Creator Fast Track, guaranteeing up to $3,000 monthly for three months.
10 sources compared
Meningitis B outbreak in Kent triggers vaccination of about 5,000 University of Kent students
32 sources compared
Nintendo Switch 2 Enables Original Switch Games to Run at 1080p in Handheld Mode
11 sources compared
Argentina Exits WHO Under President Javier Milei
23 sources compared