Notion Acquires AI Search Startup ZeroEntropy

Notion just acquired one of its own vendors, and the deal doubles as a genuinely notable career milestone for a young founder who started building deep AI infrastructure less than two years ago. Notion has acquired ZeroEntropy, a San Francisco-based AI search infrastructure startup, with the entire team joining Notion and founder Ghita Houir Alami taking over the company's newly formed Model Research team, according to Notion's own announcement of the deal.

The acquisition grew directly out of a customer relationship. Notion first got to know Houir Alami, her co-founder and CTO Nicholas Pipitone, and the ZeroEntropy team as customers of the product, and Notion AI's unified search has been running on ZeroEntropy's reranker model zerank-2 for the past two months, replacing models from larger, better-funded competitors, cutting latency by 85% in the process.

ZeroEntropy's core thesis, according to FWD Start's reporting on the acquisition, was that narrow, repeatable steps inside AI pipelines, reranking, embedding, classification, routing, and query rewriting, don't require frontier models to perform well. The company built small, task-specific models to replace general-purpose LLMs on these specific workloads, delivering results that were faster, cheaper, and often more accurate than defaulting to a large frontier model for every step.

Founded in 2024 and going through Y Combinator's Winter 2025 batch, ZeroEntropy raised a $4.2 million seed round in July 2025 led by Initialized Capital, with participation from Y Combinator, Transpose Platform, 22 Ventures, and a16z Scout, alongside angels from OpenAI, Hugging Face, and Front, according to TechCrunch's original coverage of that raise. Since launching, the platform processed close to 100 million documents and roughly 1 million queries.

Houir Alami's path is genuinely distinctive within AI infrastructure specifically. Born and raised in Morocco, she left home at 17 to study engineering in France at École Polytechnique before completing a master's in mathematics at UC Berkeley. She's spoken openly about being one of relatively few female CEOs building deep infrastructure in AI, telling TechCrunch, "There aren't many women in DevTools or AI infra, but I'd tell any young woman interested in technical problems: don't let that stop you." She has continued giving talks at high schools and universities in Morocco to encourage more young women toward STEM fields.

This acquisition connects to a broader pattern we've tracked in AI infrastructure consolidation, similar to the dynamics at play in our coverage of what a large language model requires at the pipeline level, where task-specific, cost-efficient components increasingly matter as much as the frontier model itself.

In my four years in sales at a research and advisory firm, I heard directly from CMOs and CEOs about what they wanted from AI, and search and retrieval accuracy came up constantly as a pain point for any business building internal knowledge tools or customer-facing AI products. ZeroEntropy's approach, and Notion's decision to acquire rather than just license the technology, is a useful signal that the "boring" middle layer of AI pipelines, reranking and retrieval specifically, is becoming genuinely strategic infrastructure worth owning outright.

For businesses building AI-powered search or knowledge tools internally, this acquisition is worth watching as validation that smaller, task-specific models can meaningfully outperform larger general-purpose ones on narrow, well-defined tasks, often at a fraction of the cost.

Notion acquired ZeroEntropy, an AI search infrastructure startup founded by Moroccan engineer Ghita Houir Alami, with the entire team joining Notion and Houir Alami leading a new Model Research group. ZeroEntropy's reranking model has already cut latency 85% on Notion AI's search since deployment two months ago. The startup's approach uses small, task-specific models rather than frontier LLMs for narrow AI pipeline steps like reranking and embedding.