Forget the buzz around large language models (LLMs) for a second—there’s a new player in town, and it’s focusing on a different kind of data. Meet Neuralk-AI, a French startup making waves with its AI models designed specifically for structured, tabular data. Think spreadsheets, SQL databases, or CSV files—stuff that’s neatly organized into rows and columns.

Here’s the deal: While LLMs are great for tasks like answering questions or generating text, they’re not built to handle structured data effectively. That’s where Neuralk-AI comes in. The company just announced a $4 million funding round led by Fly Ventures, with participation from SteamAI and several industry angels, including Thomas Wolf of Hugging Face and Philippe Corrot from Mirakl.

So, why does this matter? Well, structured data is the backbone of many businesses. Retailers, for example, rely on it for everything from managing product catalogs to tracking customer behavior. Neuralk-AI’s models aim to automate complex workflows, optimize product recommendations, detect fraud, and even generate sales forecasts for better inventory and pricing strategies.

The startup’s co-founder, Alexandre Pasquiou, explains, “Today’s LLMs are amazing for unstructured data, but they fall short when it comes to classic machine learning tasks based on tabular data.” Neuralk-AI plans to fill that gap by offering its models as an API, starting with commerce companies.

Over the next few months, Neuralk-AI will test its models with major French retailers like E.Leclerc and Auchan. By September, the team aims to release what it calls the “best tabular foundation model” for representation learning.

If successful, Neuralk-AI could change how businesses harness their data, proving that sometimes, the most valuable insights are hiding in plain sight—in those tidy rows and columns.