Page 631 - AI for Good Innovate for Impact
P. 631
AI for Good Innovate for Impact
Use case-2: QANDA – An AI Tutor for everyone, advancing equity
in education 4.7: Education
Country: Republic of Korea
Organization: Mathpresso Inc.
Contact Person(s):
Monsoo Jung (jason.jung@ mathpresso .com)
Donghun Oh (kilian.oh@ mathpresso .com)
Mathpresso Support Team (support@ mathpresso .com)
1 Use Case Summary Table
Category Education
The problem to be Access to quality education remains highly unequal across and within coun-
addressed tries, worsened by rapid technological advancement. Many students lack
teachers to ask questions or receive personalized guidance, limiting their
learning opportunities. This inequity obstructs progress towards enhancing
quality education, reduced inequalities, and decent work and economic
growth.
Key aspects of the QANDA provides an AI-powered learning platform that offers real-time
solution question-and-answer support through a combination of optical charac-
ter recognition (OCR) and large language models (LLMs). Beyond simple
answers, the platform delivers a personalized learning experience by analyz-
ing student behavior, curriculum data, and study patterns. QANDA’s AI
agents support not only Q&A but also note-taking, memorization, self-eval-
uation, and concept mastery to enhance self-directed learning.
Technology Generative AI (GenAI), Large Language Models (LLMs), Optical Character
keywords Recognition (OCR), Retrieval-Augmented Generation (RAG)
Data availability Private dataset
Metadata (type of Text, Image
data)
Model training QANDA utilizes large language models (LLMs) fine-tuned for educational
and fine-tuning tasks, enhanced with a retrieval-augmented generation (RAG) pipeline to
ensure factual accuracy. The platform combines optical character recog-
nition (OCR) for image-based input processing and zero-shot reasoning
for handling unseen questions. AI models continuously improve through
learning data accumulation, enabling personalized guidance and scalable
content generation across diverse subjects.
595
