---
title: README
emoji: ⛰️
colorFrom: blue
colorTo: red
sdk: static
pinned: false
---

<div align="center">

<img src="https://huggingface.co/spaces/COPA-AI/README/resolve/main/assets/banner-dark.svg" alt="COPA — Decision Intelligence" width="100%">

<a href="https://copa.team"><img src="https://img.shields.io/badge/copa.team-C4001A?style=for-the-badge&logo=safari&logoColor=white" alt="Website"></a>
<a href="https://github.com/COPATeam"><img src="https://img.shields.io/badge/COPATeam-0A0A0A?style=for-the-badge&logo=github&logoColor=white" alt="GitHub"></a>
<a href="https://linkedin.com/company/copateam"><img src="https://img.shields.io/badge/copateam-0A66C2?style=for-the-badge&logo=linkedin&logoColor=white" alt="LinkedIn"></a>
<a href="mailto:info@copa.team"><img src="https://img.shields.io/badge/info%40copa.team-6b6963?style=for-the-badge&logo=minutemailer&logoColor=white" alt="Contact"></a>

<br>

<b>COPA turns state-of-the-art AI capabilities into tailor-made decision
intelligence products</b> that change how organizations deliver services,
manage knowledge, and make decisions — one adaptable intelligence
architecture, configured for strategic, institutional, regulatory,
commercial, and operational decision environments.

<i>Intellectual rigor · practical usefulness · responsible innovation · long-term autonomy</i>

</div>

---

## 🔬 Open research — an LLM ecosystem for Armenian

A complete, auditable LLM ecosystem for Armenian, accompanying the paper
[***From Zero to Hero: An Open LLM Ecosystem for Armenian***](https://arxiv.org/abs/2609.03350) (arXiv:2609.03350):

| | Artifact | What it is |
|---|---|---|
| 📰 | [**ArmWeb**](https://huggingface.co/datasets/COPA-AI/armweb) | 4.37M-document (3.3B Gemma-token) curated Armenian news corpus, deduplicated and benchmark-decontaminated |
| 🧮 | [**ArmSTEM**](https://huggingface.co/datasets/COPA-AI/armstem) | 373K verified parallel EN–HY math and science problems (324K with step-by-step solutions) |
| 🤖 | [**arm-gemma-e4b**](https://huggingface.co/COPA-AI/arm-gemma-e4b) | Gemma-4-E4B adapted to Armenian — first open Armenian LLM with its complete training corpus and recipe |
| 🛠️ | [**Reproduction code**](https://github.com/COPATeam/armenian_llm_ecosystem) | Full pipeline, training recipes, and evaluation harness (MIT) |

Everything lives in the
[**Armenian LLM Ecosystem collection**](https://huggingface.co/collections/COPA-AI/armenian-llm-ecosystem-6a98b17dab8c0e49af0add0d).

## 🏛️ Projects

- [**Parlour**](https://parlour.copa.team) — parliamentary intelligence platform.

## 👥 Team

**Founders** — Vahan Martirosyan (Strategy & Delivery) · Henrik Sergoyan
(Operations & Partnerships) · Erik Arakelyan (Technical Architecture & AI
Systems)

**Team** — Khatun Avetisyan · Meri Davtyan · Heghine Grigoryan ·
Nane Khachatryan · Hayk Shahsuvaryan

<div align="center">
<sub>From decision to working pilot in two weeks · <a href="https://copa.team">copa.team</a></sub>
</div>