gimmy256/UgandaEdu-5K

---
license: apache-2.0
language:
- en
- lg
tags:
- education
- uganda
- curriculum
- vocational
- gemma
- elimuai
size_categories:
- 1K<n<10K
task_categories:
- question-answering
- text-generation
pretty_name: UgandaEdu-4K
---

# UgandaEdu-4K

**Training dataset for ElimuAI** — an offline-first AI education assistant
built on Gemma 4, designed for Ugandan teachers and out-of-school youth.

Created for the **Gemma 4 Good Hackathon** (April–May 2026).

## Dataset Summary

| Split | Records |
|---|---|
| curriculum_qa | 1287 |
| vocational_steps | 210 |
| lesson_plan | 206 |
| **Total** | **1703** |

## Source Documents

- **Uganda NCDC Primary & Secondary Syllabuses** — Official curriculum documents
from the National Curriculum Development Centre (ncdc.co.ug)
- **UBTEB/UVQF Vocational Frameworks** — Uganda Business and Technical
Examinations Board trade curricula (Textile & Garment Construction)

## Schema

### curriculum_qa.jsonl
```json
{
"id": "string",
"subject": "Mathematics | English | Science | ...",
"level": "Primary 5 | O-Level | ...",
"prompt": "question aligned to NCDC curriculum",
"response": "accurate, grounded answer",
"difficulty": "easy | medium | hard",
"local_context": true
}
```

### vocational_steps.jsonl
```json
{
"id": "string",
"trade": "Textile | Garment | Mechanics | ...",
"prompt": "How do I [practical task]?",
"response": "Step 1: ...\nStep 2: ...",
"tools_needed": "string",
"visual_check": "string",
"has_image_context": true
}
```

### lesson_plan.jsonl
```json
{
"id": "string",
"subject": "string",
"level": "string",
"prompt": "teacher natural language request",
"response": {
"title": "...",
"objectives": ["..."],
"materials": ["chalk", "blackboard"],
"introduction": "...",
"main_activity": "...",
"assessment": "...",
"local_context_note": "..."
}
}
```

## Usage

```python
from datasets import load_dataset

# Load all splits
qa = load_dataset("gimmy256/UgandaEdu-5K", data_files="curriculum_qa.jsonl")
voc = load_dataset("gimmy256/UgandaEdu-5K", data_files="vocational_steps.jsonl")
plans = load_dataset("gimmy256/UgandaEdu-5K", data_files="lesson_plan.jsonl")
```

## Model
Fine-tuned Gemma 4 E4B weights: [gimmy256/elimuai-gemma4-e4b](https://huggingface.co/gimmy256/elimuai-gemma4-e4b) *(coming soon)*

## Citation
```
@misc{ugandaedu4k2026,
title = {UgandaEdu-4K: A Uganda Curriculum Training Dataset for ElimuAI},
author = {Crane AI Labs},
year = {2026},
url = {https://huggingface.co/gimmy256/UgandaEdu-5K}
}
```

## License
Apache 2.0. Source curriculum documents remain property of NCDC Uganda and UBTEB.
Generated Q&A pairs are released under Apache 2.0.

Generated: 2026-04-05