Neuralk-AI Creates Tailored AI Solutions for Structured Data Analysis
Tabular data pertains to systematically arranged information displayed in rows and columns, commonly found in SQL databases, spreadsheets, or .CSV files.
Despite significant progress in artificial intelligence for handling unstructured and sequential data, large language models (LLMs) still fall short in terms of accuracy. They function by processing input tokens to generate coherent outputs instead of adhering to a specific structure. Moreover, the leading LLMs can come at a high cost, whether used via an API or operated within one’s own cloud infrastructure.
However, many organizations have adopted data strategies that utilize a data warehouse or data lake to centralize crucial information, along with data scientists who analyze this data to enhance the company’s strategies.
The French startup Neuralk-AI focuses on artificial intelligence models designed specifically for tabular data and has recently secured $4 million in funding.
“The most valuable data for companies is that which was identified long ago, organized in a table format, and employed by data scientists to create their machine learning algorithms,” remarked Neuralk-AI co-founder and Chief Scientist Officer Alexandre Pasquiou during an interview with TechCrunch.
Neuralk-AI perceives a distinctive opportunity to advance AI model development by concentrating on structured data. Initially, the company intends to provide its model as an API for data scientists in the retail industry, where data plays a critical role—addressing aspects like product catalogs, customer information, and shopping cart behaviors.
“While LLMs excel in search functions, user engagement, and addressing inquiries about unstructured documents, they struggle when reverting to classical machine learning reliant on traditional tabular data,” clarified Pasquiou.
With Neuralk-AI’s technology, retailers can enhance complex data workflows through smart deduplication and enrichment. Additionally, their models can assist in fraud detection, enrich product recommendations, and predict sales related to inventory management and pricing strategies.
Fly Ventures led the $4 million funding round, with contributions from SteamAI. A number of angel investors also participated, including Thomas Wolf from Hugging Face, Charles Gorintin from Alan, and Philippe Corrot and Nagi Letaifa from Mirakl.
The team is actively upgrading its models and plans to begin testing with major French retailers and commerce startups, such as E.Leclerc, Auchan, Mirakl, and Lucky Cart.
“In the next three to four months, we expect to roll out the first version of our model along with a public benchmark to assess its performance against industry benchmarks,” noted Pasquiou. “By September, we aspire to establish ourselves as the premier tabular foundation model in representation learning.”