--- title: Wallapop Deploys Real-Time AI Discovery From Swiss Startup Albatross Across Its Marketplace description: Spain’s largest C2C marketplace replaces its recommendation engine with AI that tracks user intent within each browsing session, delivering triple-digit engagement gains sustained over four months. author: Darie Nani (Editor-in-Chief) date: 2026-02-12T16:12:53.000Z updated: 2026-02-26T18:01:31.471Z canonical: https://www.sovereignmagazine.com/article/wallapop-deploys-real-time-ai-discovery-from-swiss-startup-albatross-across-its-marketplace image: https://cdn.nanimediahouse.com/founders-albatross.webp categories: Startups content_type: Spotlight region: Spain publication: Sovereign Magazine about: - type: Organization name: Albatross AI A.G. description: Albatross is pioneering real-time discovery with the first infrastructure for sequential embedding models trained on live events. Founded in 2024 by former Amazon AI leaders Dr. Kevin Kahn and Dr. Matteo Ruffini, together with serial entrepreneur Johan Boissard, Albatross is headquartered in Baar (Zug) with an office in Zurich and a team of 14. The company has raised $16 million from MMC Ventures, Redalpine, and Daphni. Find out more at usealbatross.a url: https://usealbatross.ai/ sameAs: - https://www.linkedin.com/company/albatross-ai/ --- Wallapop, Spain’s largest consumer-to-consumer marketplace, has deployed real-time AI discovery technology from Swiss startup Albatross across its platform. The partnership, announced on 12 February 2026, replaces the conventional recommendation system with one that interprets user behaviour within each browsing session rather than relying on historical data. **About Albatross AI A.G.** Albatross is pioneering real-time discovery with the first infrastructure for sequential embedding models trained on live events. Founded in 2024 by former Amazon AI leaders Dr. Kevin Kahn and Dr. Matteo Ruffini, together with serial entrepreneur Johan Boissard, Albatross is headquartered in Baar (Zug) with an office in Zurich and a team of 14. The company has raised $16 million from MMC Ventures, Redalpine, and Daphni. Find out more at usealbatross.a [Website](https://usealbatross.ai/) The deployment follows a four-week A/B test on 10% of Wallapop’s traffic. User engagement increased by 119%, favourites and interactions rose 105%, and purchase intentions grew 47%. Those gains held steady across a subsequent four-month production period. ## How the system works in practice Standard recommendation engines operate on batch-trained models. They track what a user has previously searched for or purchased, then serve variations of the same. If a user browses a second-hand sofa, the system continues to surface sofas. Albatross replaces that with sequential embedding models trained on live events. Every interaction within a session updates the system’s understanding of what the user is looking for. If that same user moves from sofas to rugs to floor lamps, the system recognises the shift from evaluating a single item to furnishing a room and surfaces complementary products across categories, including listings the user never searched for. The difference matters more for second-hand marketplaces than for conventional retail. Wallapop’s catalogue contains millions of unique, one-of-a-kind items with inconsistent metadata. Popularity-based and similarity-based recommendation methods struggle with that kind of inventory because the same item rarely appears twice. Albatross’s model learns from sequences of user-item interactions rather than catalogue metadata, representing products based on how buyers perceive them rather than how sellers tag them. ‘At Wallapop, we are moving toward a system that understands what users want in real time, helping buyers find the right items faster while giving sellers more effective visibility for their listings,’ said Rob Cassedy, CEO of Wallapop, in the [announcement](https://www.globenewswire.com/news-release/2026/02/12/3236917/0/en/Wallapop-and-Albatross-Sign-Strategic-Partnership-to-Bring-Real-Time-AI-Discovery-to-the-Future-of-Consumer-to-Consumer-Commerce.html). ## Unlocking supply that popularity algorithms bury Beyond buyer relevance, the system addresses a structural problem in C2C marketplaces: quality listings that disappear in volume. Traditional recommenders amplify already-popular items, which means a well-photographed listing from a power seller gets shown repeatedly while a comparable item from a casual seller never surfaces. Albatross’s system matched live buyer intent with relevant shippable listings, surfacing previously unseen items to qualified buyers. For a marketplace processing over 100 million listings annually across 21 million monthly users, that redistribution of visibility has direct commercial value. Sellers get found without paying for promotion, and buyers encounter listings they would otherwise have missed. ## Albatross and the case for European AI infrastructure Albatross was founded in 2024 by Dr Kevin Kahn and Dr Matteo Ruffini, both former Amazon AI leaders, alongside serial entrepreneur Johan Boissard. The company is headquartered in Baar (Zug) with offices in Zurich and a team of 14. It has raised $16 million from MMC Ventures, Redalpine and Daphni, including a [$12.5 million seed round](https://www.sovereignmagazine.com/article/ex-amazon-ai-leaders-launch-albatross-to-fix-what-amazon-can-t) in November 2025. The company’s platform processes over one billion events per month, updates embeddings more than 4,000 times per second and delivers predictions in under 100 milliseconds. Its research on cold-start discovery was presented at RecSys 2025 and now powers the production platform. Forbes [described](https://www.forbes.com/sites/charliefink/2025/11/18/albatross-raises-1225-million-to-build-the-second-pillar-of-ai/) the company’s approach as ‘the second pillar of AI’, distinguishing real-time perception from the generative AI models that dominate current investment. The Wallapop deal represents one of the earliest large-scale commercial deployments of adaptive, in-session AI discovery in recommerce. Integration took under seven weeks from contract to production. ## Naver’s acquisition adds scale South Korean tech group Naver completed its full acquisition of Wallapop on 30 January 2026 in a deal valued at €600 million. Naver had held a 29.5% stake since 2021 before acquiring the remaining 70.5% for €377 million. Naver operates a portfolio of C2C platforms globally, including Poshmark in the United States and Kream in South Korea. The company has stated its intention to deploy its search, advertising, payments and AI technology across Wallapop, using Barcelona as a hub for European expansion. Forbes [noted](https://www.forbes.com/sites/davidprosser/2026/02/12/how-a-swiss-start-up-is-on-its-way-to-reinventing-product-discovery/) on 12 February that the Naver acquisition could give Albatross a pathway to international deployment across that portfolio. Wallapop reported revenue of €101 million in 2024, up 13% year-on-year, and reached profitability in Spain for the first time. The company connects 21 million monthly users across Spain, Italy and Portugal. ## Further Context **Q: How does AI personalisation work?** Most AI personalisation systems use a customer’s browsing and purchase history to build a static profile, then serve recommendations based on past behaviour. Albatross takes a different approach: its sequential embedding models train directly on live events within each session. Every click, search and interaction updates the system’s understanding of what a user wants in that moment. If a user shifts from browsing sofas to looking at rugs and floor lamps, the model recognises the change in intent and adjusts recommendations within milliseconds rather than waiting for the next batch-training cycle. The system processes over one billion events per month and delivers predictions in under 100 milliseconds. **Q: Is the second-hand market growing?** The European second-hand market reached a turnover of €21.6 billion in 2024, according to a report cited by Amazon and industry analysts. BCG estimates the global resale market is growing at 10% per year and could reach $360 billion by 2030. In Spain, second-hand trade accounts for €13.8 billion annually (0.86% of national GDP), according to Wallapop’s La Red del Cambio report produced with business school ISDI. Wallapop itself reported revenue of €101 million in 2024, up 13% year-on-year, and reached profitability in Spain for the first time. **Q: What is the difference between ecommerce and recommerce?** Ecommerce deals primarily in new products with standardised catalogue data (identical SKUs, consistent metadata, predictable stock levels). Recommerce (reverse commerce) involves the resale of previously owned items, where every listing is unique, product descriptions vary in quality, and inventory changes constantly as items sell and new ones appear. This distinction matters for discovery technology: conventional recommendation engines built for ecommerce rely on product similarity and popularity signals that work when thousands of the same item exist. In recommerce, those methods fail because the same item rarely appears twice. Albatross’s system addresses this by learning from sequences of user interactions rather than catalogue metadata, representing products based on how buyers perceive them rather than how sellers tag them.