Predictive AI for the enterprise
Enterprise data is tables, not paragraphs or images. Until now, no foundation model was built purposefully for it. We did.
Neuralk was co-founded in 2023 by Alexandre Pasquiou and Antoine Moissenot — two engineers who shared a conviction that AI’s impact on business data was barely scratching the surface.
Together they have assembled a team of AI scientists and engineers with a singular focus: building AI that natively understands tabular data, rather than forcing text-based architectures onto a fundamentally different problem.
Key milestones
Bringing predictive AI to the enterprise
Neuralk’s purpose is to enable any company to predict anything.
Predict
Revenue
Today, too many AI and data projects are slow to deliver and dependent on scarce technical experts. Predictive projects are static and one-off; stitched together across silos, hyper-customized, and kept alive by the same specialists who built them.
We build Predictive AI for the Enterprise: We bring fast, secure, reliable predictions everywhere decisions are made, without implementation debt or technical barriers.
Predictions shouldn’t be handcrafted and babysat, stuck behind complex setups and expertise. They should be like electricity: always on, trustworthy, and available to anyone in the enterprise.
A world where predictions on business data are radically simpler and better.
Prediction workflows currently require many steps, from data collection and processing to feature engineering, model training, deployment, monitoring, and retraining, not to mention governance and human-in-the-loop. Work is fragmented across silos and each pipeline becomes a one-off craft project: custom-built and kept alive by the same specialists who built it.
The issue is less about training models nor making predictions in a production setting, but more about the complexity of building, deploying and maintaining production-ready prediction workflows over time, at scale.
We want enterprises to move away from traditional static prediction workflows towards dynamic agentic + foundational model workflows:
Foundational tabular models
Pre-trained predictive models that do not require re-training on client’s data, that generalize across datasets and tasks, automating training and model tuning.
Agentic workflows
Autonomous agents that can analyze the data provided, clean and enrich it, generate features, and use the foundational model to generate a prediction. The agent can monitor data changes and modify or ‘heal’ the workflow autonomously to maintain or improve predictions, adapting as reality changes. Changes are transparently shared and documented; performance and scoring are continuously monitored and historicized for a human to monitor.
We are building the gold standard for Tabular Foundation Models and putting autonomous data science workflows into every data professional’s hands.
At the core of what we do are our proprietary models: pretrained on billions of datasets, delivering instant predictions for classification, regression and timeseries — no fine-tuning required.
Our values
Scientific Rigor, Real-World Impact
We build on deep research but measure success by business outcomes. Every line of research serves a production use case.
// V. 1
Radical Simplicity
Predictive AI should be invisible infrastructure, not a perpetual project. We obsess over removing complexity.
// V. 2
Open Collaboration
We contribute to the community through open benchmarks (TabBench), publications, and partnerships. Progress in Predictive AI benefits everyone.
// V. 3
Enterprise Trust
Stateless, process-and-discard data model. Our solutions deploy wherever your data lives. No compromises on privacy and governance.
// V. 4
Speed and Ambition
From question to prediction in under a minute. Urgency is a feature, not a bug.
// V. 5
$4M in seed funding
Fly Ventures
Berlin, seed-stage enterprise & deeptech
StemAI
AI-specialized investor
Station F
World's largest startup campus
Notable angels
Thomas Wolf (Hugging Face), Philippe Corrot (Mirakl), Vincent Luciani (Artefact), Xavier Perret, Jean-Louis Quéguiner, Nagi Letaifa, Saturnin Pugnet, Julien Launay
Recognition & Partners
Meta × Hugging Face × Scaleway
1 of 5 European AI startups selected
Station F Future 40 (2024)
Top 4% of 1,000+ startups worldwide
Nvidia
Startup ecosystem partner
Microsoft for Startups
Partner program
AWS Startups Scaleway
Partner program
Google Cloud for Startups
Partner program