Add logo and shorten the org card

README.md

@@ -6,46 +6,20 @@ sdk: static
 pinned: false
 ---
 
-# Aliac
+<p align="center">
+  <img src="https://huggingface.co/spaces/AliacAI/README/resolve/main/logo-light-theme.png" alt="Aliac" width="320" class="block dark:hidden">
+  <img src="https://huggingface.co/spaces/AliacAI/README/resolve/main/logo-dark-theme.png" alt="Aliac" width="320" class="hidden dark:block">
+</p>
 
-**Production AI for European industry, built to run on your own hardware.**
+AI engineering company based in Greece. We build machine learning systems for
+industrial use, usually deployed on the client's own hardware.
 
-Aliac is an AI engineering and MLOps company based in Greece. We build
-reliable machine learning systems for industrial and enterprise teams: models
-that run on premise, pipelines that hold up in production, and evaluations
-that state their limits.
+Work areas: speech recognition, document processing, simulation surrogates,
+computer vision, and MLOps.
 
-[Website](https://aliac.eu) | [Book a call](https://cal.com/aliac/15min) |
-[LinkedIn](https://linkedin.com/company/aliac-ai) |
-[GitHub](https://github.com/aliac-ai) | contact@aliac.eu
+Models will be released here with their papers.
 
-## What we work on
-
-| Area | What we build |
-| --- | --- |
-| Speech recognition | Compact speech models that run offline on a laptop CPU, for dictation and transcription |
-| Document intelligence | Parsing and field extraction for business documents, with page-level provenance, on premise |
-| Engineering surrogates | Graph neural networks and surrogate models that replace hour-long simulations with predictions in seconds |
-| Computer vision | Counting and tracking for warehouses and production lines |
-| MLOps | Reproducible training, verified data pipelines, and quantized deployment on cloud or on-premise hardware |
-
-## Open releases
-
-Our first open models are in preparation. Each will be published with a paper
-that documents its data, training protocol, and evaluation.
-
-## How we publish
-
-- Each model comes with a paper and a model card that state its data,
-  protocol, and error rates with confidence intervals, including where it
-  falls short.
-- Each model ships with a quantized build that runs without a GPU or a cloud
-  account.
-- Each data source is listed with its licence. We publish only data we may
-  redistribute.
-
-## Work with us
-
-Aliac adapts models to your domain and deploys them on your infrastructure.
-Most engagements start with a ten-day
-[AI Readiness Assessment](https://aliac.eu/assessment).
+[aliac.eu](https://aliac.eu) |
+[contact@aliac.eu](mailto:contact@aliac.eu) |
+[GitHub](https://github.com/aliac-ai) |
+[LinkedIn](https://linkedin.com/company/aliac-ai)