Wilma AI

Wilma AI: Your digital study buddy that joins your Zoom classes, takes smart notes while you focus on learning, and sends them straight to your inbox.

Comment

About Wilma AI

Inspiration

How we created Wilma AI during a particularly rough semester at U of A. I was sitting in my Calculus II class, trying to write down everything important while also understanding complex concepts, and failing miserably at both. Looking at my notebook later that night, I found incomplete sentences, missing steps in proofs, and notes that made absolutely no sense to me just hours after writing them.

I've always been a terrible note-taker. My handwriting is a disaster, I miss key points while writing down others, and I can never organize information in a logical way. But the real breaking point came when I had to watch six one-hour lecture recordings before an exam. That's when I thought, "There has to be a better way than rewatching all these videos or deciphering my chicken scratch notes."

What We Built

Wilma AI is like having a perfect note-taking friend in every class. It:

  • Joins your Zoom lectures automatically (no need to remember to hit record)
  • Creates detailed, organized notes while you focus on actually understanding the material
  • Delivers clean, structured PDFs directly to your email
  • Saves hours of rewatching lecture recordings for exam prep

What makes Wilma special is that it doesn't just transcribe - it understands. The notes are organized by topic with clear headings, important equations and definitions are highlighted, and there are even practice questions to test your understanding. One student who tested it said it turned their three-hour review sessions into focused 45-minute study time.

How We Built It

Building Wilma meant combining several technologies to create something seamless:

  • I started with the Fireflies API and ZoomAPI for Zoom integration, which handles joining meetings and recording them reliably. The Node.js backend manages all the moving parts and ensures everything flows smoothly from recording to delivery.
  • The real breakthrough came from executing the OpenAI pipeline. It takes messy transcripts (with all the "ums," tangents, and disorganized explanations) and transforms them into clear, structured notes. This was a process of constant refinement - testing with real lectures from different subjects to make sure it could handle everything from humanities discussions to technical engineering explanations.
  • The interface is deliberately simple - just paste a Zoom link, add the class name, and click. Students already have enough complicated tools to deal with.

Challenges We Faced

Integrating AI and actually passing the generated transcripts to it to get study notes in PDF format was much harder than expected. The data flow between different services had to be carefully managed to maintain context and formatting throughout the process.

We also ran into API limits with services such as Recall AI and Fireflies which caused unexpected failures during peak usage times. I had to implement robust queuing and retry mechanisms to handle these limitations gracefully.

What We Learned

This project gave us real-world experience with API integration—the messy kind, not the textbook version. We learned firsthand how meeting bots actually behave, with all their unexpected edge cases and failures.

Debugging integrations between incompatible services became second nature, and moving clean data between APIs was way harder than we expected.

But the biggest lesson was teamwork. At first, we worked in silos, which didn’t work at all. Real progress happened when we started communicating constantly, sharing problems, and trusting each other. When our system kept breaking at 2 AM before demo day, it wasn’t individual brilliance that saved us—it was the team.

What's Next

  • Subject-specific formatting (special templates for math, science, humanities, etc.)
  • Integration with D2L to automatically organize notes by course
  • A feature to generate flashcards from the notes for quick review
  • Options to combine notes from multiple lectures into comprehensive study guides before exams

Students keep telling me the same thing: "This tool just saved me hours of rewatching lectures and deciphering my bad notes." That's exactly what I built it for, and I'm just getting started.

Built With

Share this project:

Updates