Seltz Lands $12.5 Million to Build Search Infrastructure for AI Agents

Why it matters

Search is becoming agent infrastructure, not just a consumer interface. Seltz's seed round shows investors are still willing to fund costly, full-stack web indexes if they can reduce AI labs' dependence on Google, Bing, and wrapper APIs.

Antonio Mallia's Seltz raised $12.5 million in seed funding to build web search infrastructure for AI agents, Fortune reported Wednesday.

The round was led by Speedinvest and B Capital, with participation from Italian Founders Fund, United Ventures, and Future Back Ventures, Fortune reported. In Seltz's own funding announcement, Mallia listed a broader syndicate that also includes futurepresent, Arc Investors, Vento Ventures, Mango Capital, 2100 Ventures, and angels from Google, Ramp, Tako, and Hugging Face. Seltz did not disclose a valuation.

Mallia is not pitching Seltz as another answer engine. The bet is lower in the stack: that agents and chatbots need a search system built for machines that issue long queries, run many of them in parallel, and need structured evidence from the live web rather than a ranked page of links.

"The old search methods don't work because they were architected for humans," Mallia told Fortune. His argument is that the useful evidence for an agent often sits beyond the snippet, inside body text, tables, images, and other page-level material that a human-oriented search results page was not designed to expose.

That thesis tracks Mallia's career. He studied information retrieval at the University of Pisa before earning a PhD in computer science at NYU, worked as an applied scientist on Amazon's artificial general intelligence team, and later worked as a research scientist at Pinecone. In an April post explaining why Seltz was built, Mallia wrote that at Amazon his team built the web search engine that powered Alexa's question answering, where the consumer of search results was already a machine rather than a person choosing among links.

Seltz says it owns the crawler, index, retrieval models, and ranking stack, rather than routing queries through Google, Bing, Brave, or another search provider. That is the expensive part of the plan. In the funding post, Mallia called open-web scale search "one of the most capital-intensive problems in software" and said Seltz raised the round to scale toward "10s of billions of documents" and build the go-to-market team around the product.

The product began with a news index. Seltz says the system has shipped in eight months and, on its own Dynamic News Search Benchmark, answered with 89% accuracy while returning results in under 250 milliseconds. Those benchmark results are company-reported, not an independent measurement. Fortune reported that Seltz crawls hundreds of millions of pages a day and returns results in under 200 milliseconds, with Mallia describing the product as scoring passages and extracting the exact table, text, or image an agent needs.

The gap Seltz is attacking is real, but so is the capital asymmetry. Parallel Web Systems, founded by former Twitter CEO Parag Agrawal, raised a $100 million Series B at a $2 billion valuation in April, TechCrunch reported. Exa has also raised heavily for AI-native search, and Nebius agreed in February to acquire Tavily to add agentic search to its AI cloud platform, according to a Nebius announcement.

Seltz is entering that market with a small team. Fortune reported Seltz was incorporated in the U.S., founded last October, and has 15 people, only about half a dozen full time. Mallia said the remote team is split between the San Francisco Bay Area and hubs near universities in Pisa and Leipzig, with many holding PhDs in information retrieval and several coming from Amazon's AI efforts.

The seed round buys Seltz time to prove that independent search infrastructure is not just technically cleaner, but commercially necessary. Seltz says it has a foundational lab under contract and is running pilots with teams building agentic workflows. Mallia also said the money will go into engineering, hiring, and the beginning of enterprise sales.

That enterprise push is where the argument will be tested. Developers already have access to search APIs, scraping tools, browser automation, retrieval systems, and model-provider browsing features. Seltz has to show that owning the index produces different enough documents, lower enough latency, and better enough agent performance to justify becoming another infrastructure dependency.

The strategic point is not whether Seltz can out-Google Google for people. Mallia is betting that the next search customer is not a person at all, and that the infrastructure serving that customer will be judged on freshness, provenance, latency, and machine-readable context rather than blue-link ranking.