How It Works
From idea to lesson in minutes
Configure your lesson, let the AI generate multimedia content, then review and refine before sharing with learners.
Configure
Select the subject, grade level, topic, and content style aligned to the Tanzanian curriculum.
Generate
AI produces the lesson text, images, and narration in Kiswahili within seconds.
Review
Edit, refine, and download or share the final lesson with your class.
Overview
System architecture overview
End-to-end workflow from curriculum knowledge base to multimedia content
Core Components
Three specialized AI modules
Swahili Text Generation
RAG architecture grounded in 1,200+ pages of official TIE curriculum materials.
- Semantic search over curriculum knowledge base
- Structured lesson format with exercises
- Age-appropriate Kiswahili vocabulary
Image Generation
LoRA fine-tuned diffusion models trained on 500+ culturally-relevant local images.
- Local flora, fauna, and classroom scenarios
- Science diagrams with Kiswahili labels
- Culturally appropriate visual contexts
Audio Narration
VITS-based neural TTS achieving 4.5/5 Mean Opinion Score.
- Natural prosody and intonation
- Correct technical term pronunciation
- Suitable for early-grade learners
Performance
Optimized for efficiency
Runtime Performance
NVIDIA T4 GPU cloud server
Curriculum Knowledge Base
1,200+ pages from TIE materials
Tech Stack
Built on modern infrastructure
Backend
AI Models
Infrastructure
Research
Research & development
This system architecture is based on research conducted at the University of Dar es Salaam, advancing AI applications for educational equity in low-resource language contexts.