🌍 Early Childhood Education – AI Research Hub

Welcome to the official repository hub for privacy-preserving Artificial Intelligence in early childhood education.

🎯 Our Mission > We develop specialized, locally deployable Large Language Models (LLMs) for the highly sensitive field of kindergarten and elementary pedagogy. Our goal is to enable state-of-the-art educational documentation without sacrificing data sovereignty (Zero-Cloud Guarantee).

🤝 More About Kita Digital & EleMo

Digital Educational Sovereignty: Why Kindergarten Documentation Needs its Own Local AI

Early childhood education is caught in a dilemma: pedagogical professionals are drowning in documentation duties, while lacking time for what truly matters – working with the children. Artificial Intelligence could solve this problem, but the use of conventional, cloud-based AI models is completely out of the question in this sensitive environment.

We must not upload the most intimate developmental data of young children to US servers or release it as training material for commercial tech giants. The answer to the shortage of skilled workers cannot be the sell-out of our data. That is why we are working on an alternative: Local, data-sovereign AI for elementary pedagogy.

The End of Cloud Compulsion: The "Zero-Cloud Guarantee"

The vision of KI-Insel is to build an infrastructure where state-of-the-art technology and uncompromising data protection go hand in hand.

We rely on the principle of data sovereignty through local execution. AI models in early childhood education must run directly on-site, on local hardware, or in strictly isolated, sovereign environments. No cloud connection, no API transmission to third parties, no compromises. What happens in the kindergarten stays in the kindergarten.

In Development: EleMo (The Elementary Pedagogical Model)

A technical foundation alone, however, is not enough. A general language model does not know how to document professionally and pedagogically. It tends to make judgments, fabrications, and inappropriate diagnostics.

For this reason, we are currently training EleMo, a highly specialized, fine-tuned Large Language Model (LLM) tailored exactly to the requirements of early childhood education. EleMo is trained to generate authentic educational and learning stories from raw, bullet-pointed observation data. In doing so, the model follows strict pedagogical guardrails:

Methodology according to Margaret Carr: The output is consistently structured according to the triad: Observation – Significance – Next Steps (Opportunities).
Factual Accuracy: Every statement about learning must be based on a previously described observation. What has not been observed must not be claimed or hallucinated.
The Right Tonality: The learning story is not a dry report, not a cold analysis, and certainly no developmental diagnostics. It is written as a personal, appreciative letter directly to the child.

🌐 The Kita Digital Ecosystem

EleMo is part of a larger initiative to rethink digital education and empower pedagogical professionals. Dive deeper into our resources and community:

🎙️ Podcast: Listen to Kita Digital - Bildung von morgen where we discuss media pedagogy, the digital transformation of kindergartens, and the practical impact of AI.
📚 Upcoming Book: Currently in the works, offering a comprehensive guide on implementing local AI and digital strategies in early childhood education.
🎤 Events & Outreach: We actively share our expertise at major industry events—such as the Kita Innovation Days—and partner with educational companies to raise awareness, demystify AI, and highlight the immense practical opportunities of sovereign, local solutions.

🔬 Scientific Access & Outlook

Since EleMo embodies deep-seated methodical knowledge of elementary pedagogy, handling it requires responsibility. To prevent these specialized weights from flowing unchecked into commercial software products, the model will not be released entirely free (Open Source) on Hugging Face, but as a Gated Model.

We believe in open scientific exchange. As soon as the current fine-tuning phase is completed, we will grant researchers, universities, and scientific institutions access to the model weights upon request. Our goal is to explore, together with the academic world, how specialized, locally executed AI can sustainably increase pedagogical quality in institutions without violating children's rights.

📋 Access Application Checklist

To accelerate your approval process, please ensure your request contains:

Institutional Email: Applications from generic domains (gmail, outlook, etc.) will be automatically rejected.
Research Intent: A brief 2-3 sentence description of your research project or pedagogical use case.
Agreement: Confirmation that the weights will not be used for commercial software or shared outside your institution.

🎓 Citation

If you use EleMo or the pedagogical framework in your academic research, please cite it as follows:

APA Style: > Götz, S. (2026). EleMo: Elementary Pedagogical Model for Sovereign AI in Early Childhood Education. Hugging Face Repository. Kita Digital / KI-Insel. Available at: https://huggingface.co/Earlychildhoodeducation

The future of the digital kindergarten is secure, local, and sovereign. Technology must quietly handle the burden of documentation, giving educators back the time for what truly matters: working with the children. — Sebastian Götz | Owner of Kita Digital & Initiator of KI-Insel