LevelUp - Self learning website

LevelUp: Build a Custom University Just for You.

Comment

Inspiration

You graduate with a 4.0 GPA, you look at a job description for your dream role, and you realize... you are completely unqualified.

The problem isn't that students can't learn; it's that education is rigid. Universities provide a "one-size-fits-all" syllabus, but the job market demands niche, ever-changing skills. We asked ourselves: What if we could build a custom university for every single person?

That was the spark for LevelUp—an AI Universal Architect that bridges the gap between where you are (your raw notes/resume) and where you want to be (your dream career).

What it does

LevelUp is a dual-mode intelligence engine:

Mode A: The Academic Tutor: Upload 500 pages of messy lecture slides, and it extracts the "High-Yield" topics, generating instant cram sheets and exam simulations.

Mode B: The Career Architect: Upload your Resume and define a Career Goal (e.g., "I want to be a Data Scientist"). It performs a Synergy & Gap Analysis, ignoring what you already know and building a 2-week curriculum specifically for the skills you are missing.

How we built it

Backend: We deployed a FastAPI Python service on Render. This acts as the logic core, handling file streams and managing the Google Gemini 1.5 Flash API. We engineered a "Router System Prompt" that dynamically switches the AI's persona based on input.

Frontend: We deployed a separate Next.js web service on Render. We used Tailwind CSS for a "Cyber-Dark" aesthetic and TypeScript to ensure type safety when handling complex JSON responses.

The Connection: We connected the two independent cloud services using HTTP requests, ensuring secure communication between our visual front-end and our logic back-end.

Challenges we ran into

The "Silent Killer" (CORS): Getting our Next.js frontend to talk to our Python backend on the cloud was difficult. We faced constant 404 Not Found messages because of mismatching URLs (/analyze vs /).

Context Management: We had to figure out how to feed large context (entire resumes and course modules) into the LLM without confusing it. We solved this by using Gemini's structured input capabilities.

Accomplishments that we're proud of

Conquering the "Full Stack" Gap: Moving from localhost to the cloud is notoriously difficult. We are incredibly proud that we successfully deployed two separate services on Render and got them to communicate perfectly, overcoming complex CORS policies and "Silent Failure" network errors.

The "Smart Skip" Logic: We didn't just make a chatbot. We're proud of the logic that allows LevelUp to refuse to teach you things you already know (the "Mastered" list), which makes the AI feel like a true senior mentor rather than a generic search engine.

What we learned

  1. Connecting a frontend to a backend is easy on localhost, but deploying both to the cloud requires precise configuration of ports, hosts, and environment variables.
  2. Writing the system prompt wasn't just writing English; it was programming the AI's behavior to handle edge cases (like missing files or vague goals).
  3. We learned to debug using the Browser Console and Network Tab to find errors that weren't showing up in the UI.

What's next for LevelUp - Self learning website

1.The "Life-Sync" Scheduler (Visual Timetables) Text-based plans are hard to follow. We plan to integrate a Dynamic Gantt Chart & Calendar UI (using libraries like FullCalendar or Recharts).

What it does: Instead of saying "Phase 1: 3 Days," LevelUp will generate a visual block on your actual calendar (e.g., "Monday 9 AM - 11 AM: Study React Hooks").

Adaptability: If you miss a session, you can drag-and-drop the block to the next day, and the entire curriculum automatically shifts forward.

2.The "Deep-Dive" Resource Engine (Beyond YouTube) We are expanding our retrieval system to scrape and index diverse high-quality sources, not just videos.

Multi-Modal Learning: The AI will curate a mix of Documentation (MDN/Official Docs), Interactive Articles (Medium/Dev.to), and GitHub Repositories.

Active Recall (Quizzes): At the end of every scheduled learning block, LevelUp will generate a micro-quiz (3-5 questions) based on the specific resource you just read. You cannot unlock the next module until you pass, ensuring true mastery.

3.The "Evidence Locker" (Verified Portfolios) This is our most ambitious feature. Learning is useless without proof.

Automated Project Reviews: When a user completes a mini-project (e.g., "Build a Calculator"), they will submit their GitHub link. LevelUp’s AI will scan their code for best practices, security, and efficiency.

The "Verified" Badge: If the code passes, the user earns a "Verified Skill Badge" for that specific technology.

One-Click Portfolio: LevelUp will compile these verified projects into a shareable, public portfolio link that users can send to recruiters, proving they have practical experience, not just theoretical knowledge.

Built With

Share this project:

Updates