WarmiMIND: Quechua STEM Learning Platform

Inspiration

Many young females in Peru lack access to STEM educational resources in their native Quechua language. Existing PDFs and online content are often in Spanish, making comprehension difficult.

I personally wanted to create a platform that empowers young Quechua-speaking students with culturally relevant STEM learning. As a young Latina who has developed the official website for Warmitech, an organization that inspires leadership and runs STEM workshops in indigenous Peruvian languages (@girlup @thetechgirls), I am deeply invested in making STEM accessible and culturally grounded for underrepresented communities.

This project could be integrated into Warmitech’s broader mission, giving it more visibility and impact.

CODE: https://github.com/abigailbrionesa/warmimind

What it does

PDF Upload & Processing: Students upload PDFs directly in-browser (text-based only, temporary processing).
AI Summarization: Extracts key points into easy-to-read summaries.
Question Generation: Produces exactly 5 open-ended questions per content chunk.
Cultural Adaptation: Integrates local Andean examples and references (farming, mountains, community life).
Interactive Chatbot: Answers student questions in Quechua, step-by-step or directly.

Impact:

Preserves Quechua language and oral structure.
Makes STEM education accessible and culturally grounded.
Promotes gender equity in STEM education.

How it was built

Frontend: React / Next.js

PDF upload, chatbot interface, display of summaries and questions

Backend: Node.js or Python

PDF parsing: pdf-lib or pdfplumber
Language detection: Google Cloud Translation API (detectLanguage)
Translation: translateText
AI: Gemini AI or GPT-based for summarization, question generation, and chat
Session-based temporary memory (no database)
leveraged AI for good

Workflow:

Student uploads PDF
Backend extracts text → detects language automatically
AI summarizes → adapts culturally → generates questions
Translation API converts output to Quechua
Frontend displays summaries, questions, chat interface
Student interacts in real-time

Challenges ran into

Maintaining linguistic authenticity in Quechua while avoiding Spanish interference.
Implementing session-based AI memory without storing any long-term data.
Handling PDFs of different structures and ensuring robust text extraction.

Accomplishments that we're proud of

AI adapts STEM content culturally and linguistically, not just “translated”.
Implemented real-time, session-based chatbot in Quechua.
Preserved oral teaching style: example-first, repetition, community voice.
Delivered a judge-ready system that’s fast, reliable, and culturally grounded.
Developed a scalable framework for AI-powered education in indigenous languages.

Lessons

Culture is not just decoration because it must transform content meaningfully.
AI can preserve linguistic structures and oral rhythms while teaching STEM.
Real-time, session-based AI interactions enhance student engagement and comprehension.
Retrieval quality (chunk selection) often matters more than AI model choice for summarization and question generation.