Samsung is backing a new artificial intelligence company that says it can watch video footage so humans do not have to.
Memories AI, a startup led by former Meta researchers, has developed a platform that can analyze up to 10 million hours of video. It aims to help security firms, marketers and other businesses make sense of vast archives of footage that traditional AI models struggle to handle.
Most current video models are good at processing short clips. But they begin to break down when asked to understand longer sequences or connect ideas across many hours of content. That is the challenge Memories AI wants to fix.
“All the top AI companies are focused on building end to end models,” said cofounder Shawn Shen. “But those systems usually lose context after one or two hours of video. Human memory works differently. We wanted to replicate that broader understanding.”
Shen worked as a research scientist at Meta’s Reality Labs while earning his doctorate. His cofounder, Enmin Zhou, also worked at Meta as a machine learning engineer.
From Surveillance to Marketing Strategy
The company has raised 8 million dollars in seed funding. Susa Ventures led the round, with participation from Samsung Next, Fusion Fund, Crane Ventures, Seedcamp and Creator Ventures.
Memories AI says its platform serves two primary customer groups. Security companies can use it to scan endless hours of surveillance footage and identify potential risks. Marketing teams can use the tool to spot content trends, analyze past campaigns and even generate new videos based on what works.
Instead of sending everything to the cloud, the platform uses a custom stack to clean, compress and organize video files. It can remove noise, identify important segments, and make the entire archive searchable with natural language queries. An aggregation layer helps generate reports based on the indexed material.
The team says this approach gives companies a searchable memory system for video, not just a tool for summarizing clips.
A Bet on Privacy and Long Form AI
Samsung Next’s investment is also tied to on-device computing. By allowing video to be processed locally, Memories AI avoids many of the privacy concerns that come with cloud storage.
“One reason we were excited about this team is the potential to bring advanced video AI directly onto devices,” said Sam Campbell, a partner at Samsung Next. “That unlocks real value for consumers who are wary of putting home security footage in the cloud.”
Future updates will make syncing easier. Instead of uploading full libraries, companies will be able to connect shared drives and automatically process video in real time.
Shen envisions the platform eventually serving as a visual memory assistant. He believes it could be integrated with smart glasses, personal photo libraries, or even humanoid robots. The goal is to give machines better memory and context for real world activity.
Memories AI currently employs 15 people and is using the new funding to grow its team and improve search capabilities.
It will have to compete with other startups working on long term video memory, including Twelve Labs and Google. But Shen believes the company’s approach is more flexible and better suited to a wide range of use cases.
“It is not just about making video searchable,” Shen said. “It is about making video useful over time.”