Memories.ai Collaborates With NVIDIA To Build AI That Can See And Remember!

These days, AI seems to know a lot, especially when it comes to text. Yet the world isn’t just words on a screen. It’s full of sights, scenes, and moments we experience every day. For AI to be genuinely useful in our daily lives, it needs to do more than read or respond, and which is to see and remember what it encounters. Well, that’s exactly what Memories.ai is working on.

 

TL;DR

 

• Memories.ai builds AI that can “see and remember” for wearables and robots.
• Partners with NVIDIA and Qualcomm to power its visual memory model.
• Uses LUCI, a wearable device, to collect real-world video data.

Founded by Shawn Shen and CTO Ben Zhou, the company is building a visual memory layer for wearables and robotics. The goal is to let AI store and recall videos from the real world. “AI is already doing really well in the digital world.

"What about the physical world? AI wearables, robotics need memories as well. … Ultimately, you need AI to have visual memories. We believe in that future,” Shen said.

The idea came while the duo was developing the AI behind Meta’s Ray-Ban glasses. They realized AI would be far more useful if it could remember the videos users recorded in real life. That insight led them to spin out of Meta in 2024 and start building both the models and the infrastructure to make visual memory possible.
 
So far, Memories.ai has raised $16 million and developed its Large Visual Memory Model (LVMM). According to Shen, building this visual memory layer required two key steps: setting up the system to store and organize videos, and gathering the footage needed to train the AI effectively.

To do that, the team created LUCI, a wearable device that collects real-world videos from data collectors. The LVMM will also run on Qualcomm processors and leverages NVIDIA’s Cosmos-Reason 2 and Metropolis tools to power its memory capabilities.

Shen believes the possibilities are huge.  While wearables are the focus for now, this AI could let robots and devices see, remember, and act more like humans in the real world.