Albatross, a Zurich startup founded by former Amazon AI leaders Kevin Kahn and Matteo Ruffini alongside serial entrepreneur Johan Boissard, has raised $12.25 million and launched from stealth with what it calls “the second pillar of artificial intelligence.” Generative models can create content, but they do not understand what a user wants at the moment of creation. Albatross says it has built a model that does. The company frames it as an overlooked half of the AI stack. Its system reads behavior as it happens and adapts instantly, which they argue is the missing foundation of digital commerce.
The round was led by MMC Ventures with participation from Redalpine, Daphni, and several strategic angels. Carrefour invested through Daphni’s fund. Albatross has now raised $16 million dollars in total funding and employs fourteen people in Switzerland. Kahn said the selection of investors was about access to customers and real data rather than capital. “Money is easy. What is harder is finding partners you want to build a business with,” he said.
The founders worked together at Amazon in music, Alexa, and Prime Video. Their team explored early forms of real time signals, but the company’s legacy infrastructure made the approach difficult to scale. They left to build a version without those constraints. Their system captures every user action in a session and passes it into a transformer model that behaves like a language model for intent. The inputs are event triplets: user, action, item, instead of words. The model analyzes not just the action but the sequence of actions and the context that connects them. It updates continuously and responds in milliseconds without retraining.
The result is an API layer that controls search and discovery on websites and apps. It replaces recommendation systems based on popularity or user history. “Traditional systems cannot capture what a person is doing right now. Our system perceives and adapts instantly, so every search and feed reflects the user’s intent at that moment,” Kahn said.
The company’s two flagship products, Real Time Discovery Feed and Multimodal Search, are already in production. The platform processes two to three billion events each month and generates tens of millions of predictions. One large secondhand marketplace in Spain reported a 175% increase in engagement and more than one hundred thousand additional product acquisitions during early tests. Over half of these items had never been shown to any user before. A major travel platform in Germany is also live.
MMC Ventures conducted extensive diligence before investing, including interviews with customers and employees of existing personalization vendors. “Albatross moves beyond static algorithms and builds systems that understand context as it happens,” Mina Samaan, general partner at MMC Ventures, said in the press release. “Their architecture changes how businesses engage customers online.”
Generative AI has increased the volume of content on the internet and added complexity for retailers, marketplaces, and travel sites. Discovery has become a structural problem. Digital platforms depend on users knowing what they want or using a search box to find it. Kahn says that model no longer fits behavior. “Imagine entering Spotify or Netflix with only a search bar. You would never watch a movie. People need to see what is out there, and the web does not do that well,” he said.
Albatross sees real time perception models as the corrective force. If generative AI floods the web with content, real time AI sorts it. Their long term plan is to open parts of the platform as standalone services for event processing, embeddings, and context aware search. For now, the focus is on expanding deployments and turning pilots into multi year contracts.
Kahn shared this quote from his daughter’s presentation to her school class: “We help people find things they love that they did not know they loved.” That, he said, is their north star.