Inspiration
STEM education is in crisis — not because students lack intelligence, but because they lack the right tools to keep up.
Walk into any university today and you will find STEM students drowning in a flood of complex lecture notes, mathematical proofs, biological processes, programming concepts, and engineering principles — all competing for the same limited study time. The pressure is relentless. Exams come fast. Deadlines pile up. And the gap between what a student understands and what they need to understand keeps growing wider.
The tools available to most students have not evolved to meet this reality. They are still copying notes by hand, creating flashcards manually, and guessing which topics to prioritise the night before an exam. There is no intelligence behind any of it — just effort, anxiety, and hope.
As a STEM student myself, I lived inside this problem every single day. I watched classmates fail not because they were not working hard enough, but because they had no system. No visibility into where they were falling behind. No intelligent support to help them prepare smarter.
I built Prepify AI because every STEM student deserves access to the kind of intelligent academic support that was previously only available to those who could afford private tutors or elite institutions. The inspiration was simple — use artificial intelligence to level the playing field and give every student, regardless of background or resources, a genuine chance to succeed in STEM.
What It Does
Prepify AI is an intelligent academic productivity platform built specifically to help students study, understand, and perform better in STEM subjects. It is not a note-taking app or a simple to-do list — it is a full AI-powered academic operating system designed around how STEM students actually learn.
AI-Powered Note Summariser Students paste or upload their STEM lecture notes — whether covering calculus, computer architecture, biology, physics, or programming — and Prepify AI extracts the content, sends it to the AI, and returns a fully structured summary with headings, key concepts, definitions, and critical insights. For scanned or image-based PDFs, an automatic OCR fallback silently processes the document and extracts readable text before passing it to the AI.
Adaptive Quiz & Flashcard Generator From any lecture content, students generate custom quizzes with configurable difficulty levels, question types, and timers. After each quiz, the AI delivers immediate feedback, performance analysis, identified weak areas, and personalised study recommendations — all tailored to the specific STEM topic being studied.
Smart Study Schedule Generator Students input their available study hours, upcoming exam dates, assignment deadlines, and weakest STEM subjects. The AI analyses all of this data and generates a fully personalised daily and weekly study schedule with intelligent workload balancing — ensuring students spend more time on the subjects where they need it most.
Academic Risk Predictor Prepify AI continuously monitors assignment completion rates, quiz performance trends, study consistency, and productivity scores to calculate a risk level for each course — alerting students before academic decline in a STEM subject becomes irreversible.
AI Academic Chat Assistant A multi-turn conversational AI assistant that answers STEM questions, helps students plan for exams, generates revision questions, explains complex concepts, and provides personalised academic guidance — all informed by the student's own uploaded notes and academic history.
GPA/CGPA Performance Tracker Students track semester GPA, monitor CGPA growth over time, set academic targets, and visualise their performance trajectory across all STEM and non-STEM subjects.
Focus Session Manager Pomodoro-style focus sessions with deep work mode, exam revision mode, distraction control suggestions, productivity analytics, and study streak monitoring — designed to help STEM students maintain the concentration levels that complex subjects demand.
Multi-Channel Smart Notification System Intelligent, preference-driven notifications delivered across browser push and email via Resend API — respecting quiet hours, priority filters, and category preferences set by each student.
Academic Calendar A colour-coded monthly and weekly calendar that pulls assignment due dates, exam dates, and study sessions into one visual overview with event detail panels and completion tracking.
How We Built It
Prepify AI was built using MeDo — an AI-powered full-stack generation platform — through structured, multi-turn natural language conversations. OpenAI API powers every AI feature across the platform, and Resend API handles all email notification delivery.
Rather than engineering every component manually, MeDo served as an AI full-stack collaborator — generating, refining, and evolving the platform progressively through iterative conversation across multiple development sessions.
Using MeDo, I was able to:
- Generate a full-stack academic SaaS application with 13 pages
- Build complex multi-page dashboard architecture
- Create AI-powered STEM academic workflows
- Implement a production-grade multi-channel notification system
- Integrate the OpenAI API for note summarisation, quiz generation, study scheduling, risk analysis, and conversational chat
- Integrate the Resend API for branded HTML email notifications with dynamic data and CTA buttons
- Integrate the OCR.space API as an automatic fallback for scanned or unreadable PDF documents
- Build a structured database schema across multiple tables with full Row Level Security policies
- Deploy seven independent edge functions handling all AI processing on the backend
- Iterate rapidly through multiple versions of the platform entirely through natural language
The platform evolved from a simple assignment tracker into a full AI-powered academic operating system — entirely through conversational generation with MeDo.
Challenges We Ran Into
The PDF Extraction Problem The single biggest technical challenge in building Prepify AI had nothing to do with the AI — the OpenAI integration worked perfectly from the start. The real problem was getting readable text out of the documents students were uploading. Many STEM lecture notes and academic PDFs are scanned documents — images disguised as PDFs with no readable text layer underneath. Standard document parsing would return empty or corrupted content, leaving the AI with nothing to process. To solve this, I implemented an automatic OCR fallback system using the OCR.space API that silently detects unreadable PDFs, extracts the text from scanned pages, and passes it cleanly to the AI — all without the student ever knowing it happened.
The MeDo Production Deployment Bug One of the most frustrating challenges was a platform-level deployment issue with MeDo itself. Every AI feature and file upload worked perfectly inside the MeDo preview builder environment — but on the live published URL, file uploads were silently failing. After extensive debugging, the root cause was identified as a combination of two issues: the OpenAI API key and OCR.space API key were bound only to the development environment and not the production server, and the Supabase storage bucket had permission gaps in its Row Level Security policies that blocked authenticated users from writing files on the live site. All seven edge functions had to be redeployed after fixing the environment variables and storage policies. This challenge was particularly difficult because it consumed significant development credits on a limited free-tier plan and required precise, credit-efficient prompting to resolve without regenerating any existing features.
Building Complex AI Workflows Conversationally Designing a system where a student's uploaded notes flow through text extraction, OCR fallback, AI processing, and structured response generation — all as a seamless single action — required careful architectural thinking and significant iteration through MeDo to get the sequencing right.
Accomplishments That We're Proud Of
- Built a fully functional, multi-page AI academic platform from concept to deployment using MeDo
- Successfully integrated three real APIs — OpenAI, Resend, and OCR.space — into a single cohesive platform
- Designed and implemented a production-grade multi-channel notification system with retry logic, quiet hours, and priority filtering
- Solved a real document pipeline problem with an automatic OCR fallback that works invisibly for the student
- Built an AI that generates study schedules, quizzes, risk predictions, and academic summaries from a student's own real STEM content — with zero mock or placeholder logic
- Navigated and resolved a complex platform-level production deployment bug on MeDo with limited credits through precise, efficient prompting
- Delivered an interface inspired by Notion, Linear, and Perplexity — clean, modern, and genuinely usable on mobile
What We Learned
Building Prepify AI taught me that the hardest problems in software are rarely where you expect them.
The AI worked. The challenge was the data pipeline feeding it — and then the deployment environment running it.
The MeDo production bug was the most educational experience in this entire project. It taught me the critical difference between a development environment and a production environment, how API keys must be explicitly bound to production servers, how Row Level Security policies in Supabase behave differently in live deployments, and how to diagnose and fix infrastructure issues through precise, structured communication with an AI builder — all while conserving limited resources.
I also learned that building conversationally with MeDo is not a shortcut — it is a genuinely different and powerful way of engineering software. Describing intent clearly, refining through feedback, and iterating architecture through natural language pushed me to think more carefully about what I was building and why before any code was written.
Most importantly, I learned that AI tools like MeDo genuinely democratise software development — making it possible for a student with a real problem and a clear vision to build and deploy a production-grade platform without a traditional engineering team.
What's Next for Prepify AI
Prepify AI is built to grow — and the roadmap ahead is focused entirely on deepening its impact on STEM education specifically.
Planned next steps include:
- STEM Subject Intelligence — subject-specific AI modes for Mathematics, Physics, Chemistry, Biology, and Computer Science with formula recognition, equation explanation, and concept mapping
- Integration with university LMS platforms such as Moodle and Blackboard for automatic assignment and deadline syncing
- Collaborative study groups for peer-to-peer STEM problem solving and shared note libraries
- Voice input for note-taking and AI queries — making the platform accessible to students with disabilities
- Predictive CGPA forecasting based on quiz performance trends and study consistency data
- Gamification — streaks, badges, and academic milestones to keep STEM students motivated through long and difficult courses
- Multi-language support for students across Africa, Asia, and other underrepresented regions in global STEM education
- Laboratory simulation guides — AI-generated step-by-step practical guides for common STEM laboratory experiments
- Export to Google Calendar and Apple Calendar for seamless deadline management across devices
The vision for Prepify AI is to become the academic operating system that every STEM student in Africa and beyond relies on — to study smarter, understand deeper, and never face academic failure alone.
Built With
- api
- medo
- openai
- supabase
Log in or sign up for Devpost to join the conversation.