# Neuralk-AI

> Neuralk-AI is a Paris-based deeptech startup building foundation models for tabular, structured data - the spreadsheets, SQL tables and product catalogs that run retail and commerce but that large language models handle poorly. Its in-house model, NICL (Neuralk In-Context-Learning), delivers predictions via API for tasks like product classification, deduplication, catalog cleaning, demand forecasting and personalization, and the company maintains TabBench, an open benchmark for evaluating tabular models on real industrial use cases. Founded in 2023 by Antoine Moissenot and Alexandre Pasquiou, Neuralk-AI raised a $4M seed round in February 2025 led by Fly Ventures, with angels including Hugging Face co-founder Thomas Wolf and Alan's Charles Gorintin.

- **Founded:** 2023
- **Headquarters:** Paris, France
- **Founders:** Antoine Moissenot (Co-Founder & CEO), Alexandre Pasquiou (Co-Founder & Chief Scientific Officer)
- **Team size:** ~15 employees
- **Products:** NICL (Neuralk In-Context-Learning), Tabular Foundation Model API, Neuralk SDK, TabBench
- **Notable:** Raised a $4M seed round in February 2025 led by Fly Ventures with backing from Hugging Face co-founder Thomas Wolf., Building what it positions as the first tabular foundation model focused on commerce., Released TabBench, an open-source benchmark for tabular models on industrial use cases.

## Products & services

- **NICL (Neuralk In-Context-Learning)** — Neuralk's in-house tabular foundation model, delivering state-of-the-art performance on industrial use cases such as product classification, regression and data deduplication.
- **Tabular Foundation Model API** — API access to a representation model that ingests structured data (product catalogs, customer info, cart histories) and returns numerical representations powering predictive analytics, personalization and data-quality tasks.
- **Neuralk SDK** — A scikit-learn compatible Python interface that lets developers drop Neuralk's model into existing ML pipelines.
- **TabBench** — An open-source benchmark for evaluating machine learning models on tabular data through the lens of real-world industry use cases, comparing NICL against TabICL, TabPFNv2, XGBoost, CatBoost and LightGBM.

## Achievements

- Raised a $4M seed round in February 2025 led by Fly Ventures with backing from Hugging Face co-founder Thomas Wolf.
- Building what it positions as the first tabular foundation model focused on commerce.
- Released TabBench, an open-source benchmark for tabular models on industrial use cases.
- Published multiple tabular foundation model research papers (generative modelling, multimodal tabular learning, uncertainty benchmarking).
- Attracted an angel roster spanning Alan, Mirakl and other French tech leaders.

## Latest updates

- **2025-02** — Announced $4M seed funding led by Fly Ventures, with angels including Thomas Wolf and Charles Gorintin.
- **2025-06** — Neuralk-AI featured in technical coverage (Twinee) diving into its tabular deep-learning approach.
- **2026** — Published a series of tabular foundation model research papers and continued developing NICL and the TabBench benchmark dashboard.

## Links

- Website: https://neuralk-ai.com
- LinkedIn: http://www.linkedin.com/company/neuralk-ai
- GitHub: https://github.com/Neuralk-AI

---

Profile page: https://yespress.io/neuralk
Published by YesPress — https://yespress.io
Last updated: 2026-07-03