Customer Success Story : Neuralk-AI

"Scaleway enabled Neuralk-AI to significantly reduce the training time of its Tabular Foundation model thanks to its high-performance GPUs and scalable cloud, thereby accelerating the launch of TabBench, the first ML benchmark tailored to industrial challenges."

About the company

Neuralk-AI is developing an AI agent powered by the first Tabular Foundation Model for commerce. Unlike traditional models trained on text or images, this model is specifically designed to understand and predict from structured data, such as product catalogs, transaction logs, or customer files. It excels at interpreting the structure and complex relationships between rows and columns, even in noisy, unbalanced, or disordered datasets. This enables unmatched accuracy for critical use cases such as product recommendations, catalog optimization, and personalized marketing.

As part of their first product release, Neuralk-AI has also launched TabBench, the first benchmark specifically built to evaluate AI models on real-world, enterprise tabular data. TabBench targets practical commerce use cases such as product categorization, deduplication, and moreproviding the industry with a relevant standard to measure and compare model performance in realistic, high-impact scenarios.

Challenge

Neuralk-AI quickly encountered the technical limitations of traditional cloud solutions when training its Tabular Foundation Model. Efficient processing of massive tabular data with thousands of rows and columns requires a highly specialized GPU infrastructure capable of handling large attention matrices, heavy activation loads, and fast processing in a single pass.
Traditional approaches, including the latest ones based on in-context learning (ICL), face memory constraints and performance bottlenecks. To continue innovating, Neuralk-AI needed a cloud environment offering extended memory, seamless scalability, and reliability essential conditions for supporting its ambitions in advanced tabular AI.

Results

To meet these needs, Neuralk-AI chose Scaleway's flexible cloud infrastructure, equipped with the latest generation of GPUs with high memory capacity. This solution enabled the efficient processing of massive volumes of data and complex models, while benefiting from dynamic resource allocation, seamless scalability, reliable workload management, responsive support, and competitive pricing.

By leveraging our high-performance GPUs and scalable cloud infrastructure, Neuralk-AI reduced its model training time by more than 50%, accelerating the development of its solutions. In addition, thanks to our flexible resource allocation, we helped Neuralk-AI optimize its costs,reducing its IT expenses by a factor of 3 compared to other providers. Thanks to Scaleway, Neuralk-AI was able to accelerate its development and push the boundaries of tabular AI.

Future plans

Neuralk-AI plans to increase the number of nodes it uses in order to accelerate its roadmap by the end of the year.