Update README.md

README.md

@@ -7,4 +7,28 @@ sdk: static
 pinned: false
 ---
 
-Edit this `README.md` markdown file to author your organization card.
+## About
+
+**SynthFairCLIP** is a research initiative focused on fair vision–language models.
+
+We study how to reduce bias in CLIP-style models by combining:
+
+- **Real data** from large-scale datasets such as DataComp/CommonPool.
+- **Synthetic data** generated with state-of-the-art diffusion models.
+- **Curation and balancing** of demographic attributes across professions, activities and contexts.
+
+---
+
+### What we release
+
+- **CLIP models** trained on hybrid real–synthetic data.
+- **Large-scale WebDataset shards** of synthetic / hybrid image–text data.
+- **Eval tools and benchmarks** for analysing bias and fairness in CLIP-like models.
+
+If you use our resources, please consider citing the SynthFairCLIP project.
+
+---
+
+### Acknowledgement
+
+We acknowledge EuroHPC JU for awarding the project ID EHPC-AI-2024A02-040 access to MareNostrum 5 hosted at BSC-CNS.