Offline AI Math Tutor for Early Learners (Ages 5–9) — Adaptive, Multilingual, Low-Resource EdTech Solution | AIMS KTT Hackathon | Siltanu Alemu
# 🧠 AI Math Tutor for Early Learners (Offline, Multilingual, Adaptive)
## Overview
This project presents an offline AI-powered math tutor designed for children aged 5–9 in low-resource environments. The system delivers foundational numeracy instruction through a combination of visual representations, audio prompts, and interactive feedback, while adapting dynamically to each learner’s progress.
The tutor operates entirely offline, runs on CPU-only devices, and supports multilingual and code-switched input, making it suitable for real-world deployment in settings with limited connectivity and shared devices.
---
## Problem Context
Across many low-resource settings, children complete early primary education without mastering basic numeracy skills. By ages 8–9, many learners still struggle with:
- number recognition
- simple arithmetic operations
- problem-solving skills
The core challenge is not access to schooling, but the lack of **personalized instruction**, compounded by language transitions (e.g., moving from Kinyarwanda to English).
---
## Solution
The system delivers a complete end-to-end tutoring experience:
- presents math problems using visuals and audio
- accepts responses via text or speech (optional ASR)
- detects and adapts to the child’s language
- provides immediate feedback and guided correction
- repeats questions when needed to reinforce learning
- adapts difficulty using learner performance
The focus is on **teaching**, not just evaluation.
---
## Core Capabilities
### 1. Adaptive Learning (Knowledge Tracing)
A Bayesian Knowledge Tracing (BKT) model tracks mastery across skills:
- counting
- number sense
- addition
- subtraction
- multiplication
- division
- word problems
The system updates learner state after each interaction and selects the next item accordingly.
---
### 2. Multilingual & Code-Switched Interaction
The tutor supports: …