Update README.md

README.md

@@ -1,10 +1,87 @@
+
+**Elixir**
+
+## 🏷️ Tagline (visible sous le nom)  
+**Sovereign AI for PDF intelligence – Multimodal, local, efficient.**
+
 ---
-title: README
-emoji: 🦀
-colorFrom: pink
-colorTo: yellow
-sdk: static
-pinned: false
+
+## 📝 Description (Organization Card)
+
+**What is Elixir?**  
+Elixir builds **sovereign multimodal AI models** to extract, structure, and activate data from complex PDF documents — **locally**, with **no external dependencies**, and with **full control over your data**.
+
+We focus on **regulatory and sensitive use cases**, especially in finance, legal, and public sector — starting with recurring documents such as **KIDs**, **financial reports**, or **technical annexes**.
+
+---
+
+### ⚙️ Our Models – Small, Specialized, Sovereign
+
+We develop and fine-tune our own compact LLMs and VLMs, tailored to the needs of regulated organizations.  
+These models, collectively named **SAGE models** (Sovereign AI for Governance & Extraction), are:
+
+- Efficient enough to run on a **standard CPU or Apple chip**  
+- Specialized for **real-world document structures**  
+- Fine-tuned on **in-house datasets** we build ourselves (see: **Elixir Corpus**)
+
+We offer a selection of these models for **open use and testing** directly via Hugging Face Spaces.
+
 ---
 
-Edit this `README.md` markdown file to author your organization card.
+### 📚 Elixir Corpus – Our Data Foundation
+
+All our models are trained on the **Elixir Corpus**, a structured collection of open datasets built from public and regulatory documents.  
+Each subset focuses on a key domain: finance, public governance, legal frameworks, ESG reporting, and more.
+
+✅ PDFs (text-based or scanned)  
+✅ Tables, texts, images, and charts  
+✅ Built from **OpenData**, web scraping, and internal sources  
+✅ Annotations: manual, semi-automated, or model-assisted  
+✅ License: **Apache 2.0** (open for research & commercial use)  
+✅ Tested with models and available for download or demo
+
+First available dataset: **KIDs Dataset (Finance)**
+
+---
+
+### 🌍 Why Elixir?
+
+- 💡 **No hype** – real value from real data  
+- 🔐 **No cloud** – full data sovereignty  
+- ⚙️ **No guesswork** – structured output for real-world operations  
+- 🌱 **Low carbon** – 100x more efficient than standard cloud AI  
+- 🧠 **Open knowledge** – accelerating compliance & innovation for everyone
+
+---
+
+### 💼 Who is it for?
+
+Elixir is built for:
+
+- Researchers working on document AI, multimodality, or regulatory tech  
+- Companies and public institutions looking to embed **sovereign AI** into their infrastructure  
+- Builders who need **clean, structured data** to train or benchmark new models
+
+---
+
+### 🧪 Want to try?
+
+Use one of our models or datasets.  
+Explore a demo.  
+Clone a Space.  
+Or just reach out.
+
+We believe **your data should work for you — not the other way around**.
+
+Let’s make it happen.  
+**→ https://huggingface.co/Elixir**
+
+---
+
+Souhaites-tu que je t’aide à :
+- Ajouter un logo/banner ?
+- Générer un README type pour les datasets ?
+- Faire la page d’un modèle “SAGE” ?
+- Proposer un petit Space de démo ?
+
+Je peux tout rédiger/structurer pour toi selon tes besoins.