About
Turning raw data into trusted infrastructure.
AI changed what is possible. What remains is to build what is reliable.
Our conviction
Everyone has access to the same AI. Costs are collapsing, and open-source models are months away from matching the most powerful ones. Soon, any organization will have access to reasoning that surpasses human capability, for almost nothing. AI is no longer an advantage. It is a commodity.
Yet most organizations act as if the model were the answer. They plug an LLM into a pile of documents and hope artificial intelligence compensates for the disorder in their data. It does not work. It has never worked with any technology, and it will not work with this one.
What makes the difference is what you feed the AI: the data, the rules, the processes.
The vast majority of that knowledge lives in documents no one has ever been able to exploit at scale. AI can read them one by one. It cannot turn them into a reliable, verifiable, actionable system.
The problem is not new. ERPs solved it for transactions, CRMs for customers. But at what cost: armies of operators to enter, normalize, and link data that already existed in documents. Tools costing hundreds of thousands of euros. Months of deployment. And despite all that, 80% of enterprise data remains out of reach, locked in PDFs, emails, and specifications these systems were never built to handle.
That barrier is breaking.
Structuring directly from the source, at scale, without manual re-entry, with a level of rigor that was previously out of reach: this is now possible. And unlike raw AI tools, whose opacity prevents iterative improvement, the mechanism is transparent. Every processed document is traced, applies human business rules, and alerts on new cases or critical inconsistencies. Data and human insight converge at every iteration: processes gain precision, data becomes predictive, and time is freed for what no system will replace: expertise, human relationships, and judgment.
And this time, data stays where it belongs. The era of cloud platforms that lock organizations into proprietary ecosystems, where switching providers costs more than the product itself, is ending. Structured data belongs to those who produce it. It must remain sovereign, portable, deployable wherever the organization decides.
This is a fundamental shift in how work gets done. Teams move from manual re-entry and control to a different role: refining processes, building unique expertise, steering systems that operate on a foundation of trust.
This is the infrastructure Astora is building.
The question is no longer whether your data will be structured.
It is when, and on whose terms.
The team
Erwan Colombel
CEO
Théo Lecoutre
CTO
Gabriel Shenouda
AI Lead
Maxime
Sales
Gabin
Full-Stack Developer
Nathan
Data Engineer
Journey
2026
Expansion
Gradual expansion into new verticals: energy, public sector, aerospace. The platform proves its versatility beyond construction.
2025
Commercial launch
First deployments in the construction industry. The sector stands out for its document complexity, high stakes, and strong demand.
Winter 2024
Recognition
Academic Startup Award (Microsoft, New York and SAP, Paris). Most Promising Emerging Startup of the Year (TechInnov). BFTE Grant from BPI France. Pépite Hodéfi Award, Hauts-de-France Region.


Summer 2024
Going full-time
Astora is founded. The founding team leaves academia to focus full-time on industrializing the structuring engine.
2023
Applied research
Launch of a research project on entity extraction from unstructured documents, independent of large language models. In collaboration with the DVRC laboratory and Centrale Lille.

2022
Origins
Large-scale fact-checking project: automated cross-referencing of news publications, scientific studies, and open data to assess discrepancies and build personalized learning paths. Supported by Centrale Lille.