BadgerFlix

Inspiration

As students, we all know the pain of the 90-minute lecture that you only needed 10 minutes of… and the awkward feeling of wanting to ask a question but being too shy, tired, or anxious to speak up.

We kept hearing the same complaints from friends:

“I missed one class and now I’m lost.”
“The recordings are so long, I never actually rewatch them.”
“I don’t want to look dumb in front of 200 people.”

We asked ourselves: What if lectures felt more like Netflix, short episodes, easy to binge, and personalized, instead of one giant block of pain?

That's how we got the idea of BadgerFlix, which is Netflix but for education

What it does

BadgerFlix is built first for students.

It’s for the people who leave a lecture thinking, “I kind of get it… but not enough to pass the exam,” or open a massive PDF and immediately lose focus. Instead of forcing them to rewatch a 90-minute recording or reread dense notes, BadgerFlix lets students turn the content they already have into short, bingeable explainers.

Students can upload learning material they’re allowed to use, like:

lecture slides
PDFs / textbook pages
their own notes
lecture recordings shared by the instructor

From there, BadgerFlix uses AI to re-teach the material, not just cut it up. It:

breaks the content into short episodes organized by concept
simplifies the language and highlights key ideas
presents everything in a Netflix-style interface with subjects as “genres,” courses as “shows,” and episodes you can click and consume one by one

So when a student opens a course in BadgerFlix, they’re not just watching the same lecture chopped into pieces. They’re going through fresh, AI-generated explainer episodes based on their course content, designed to be easier to follow and easier to binge.

Inside each episode, they get: an AI tutor (WhisperChat) that answers questions about that specific episode

“Explain this like I’m in high school.”
“Give me an intuitive example.”
“Make a quick practice question for me.”

If they’re still confused even after the AI explanations, they can:

ask the instructor anonymously questions from that episode
the instructor sees anonymous questions grouped by episode

BadgerFlix takes the content students already have, turns it into short, AI-generated explainer episodes, lets them get help from an AI tutor, and gives them a safe, anonymous path to the instructor when they still need human backup.

How we built it

We split the project into three main pieces:

Frontend – “Netflix interface” 1.1. Built with Next.js + React and Tailwind CSS 1.2. Home page with:
- hero banner
- horizontal rows grouped by subject 1.3 Course page that shows episodes like a season of a show 1.4. Episode page with:
  - summary
  - key points
  - AI chat box
  - anonymous Q&A section
Backend – API + logic 2.1 Built with FastAPI (Python) 2.2 Endpoints for: - uploading lectures -listing subjects, courses, and episodes -asking the AI tutor -posting + retrieving anonymous questions - instructor answering questions - In-memory storage / simple data models (Course, Episode, Question) for hackathon speed
AI pipeline 3.1 Speech-to-text: transcribe uploaded lecture audio 3.2 LLM episode generator: take the transcript + title and generate: - episode chunks - titles - summaries - key points 3.3 AI tutor: use the episode’s content (title, summary, points, transcript excerpt) + student question to generate a focused answer

Challenges we ran into

AI model juggling. We initially mixed different APIs (OpenAI vs Gemini) and broke our backend more than once. We had to refactor and commit to a single LLM path to keep things stable.
Time vs ambition. We really wanted to generate actual videos (slides + TTS + MP4), but that was too heavy for a 24-hour hackathon. We decided to prioritize: - real end-to-end AI flow - clean UI
Prompt design. Getting the model to reliably output structured JSON for episodes, flashcards, or quizzes was trickier than expected. One bad generation would crash the parser, so we had to tighten prompts and add simple fallback handling.
Anonymous Q&A design. We had to think about:
- how to store questions
- what instructors see
- how to keep questions anonymous but still useful to everyone and still keep the implementation simple enough for a hackathon.

Accomplishments that we're proud of

We built a real, end-to-end pipeline where you can upload a lecture and instantly get it broken into structured episodes, open any episode and ask an AI tutor about that specific chunk, and even ask a question anonymously so the answer shows up for the whole class. The app actually feels like Netflix, not a boring LMS: dark theme, horizontal rows, hover effects, and a clear “Season / Episodes” vibe. We didn’t just build another chatbot, we used AI where it actually makes sense: to structure content into episodes, to tutor inside each episode, and to generate study aids, instead of just slapping a chatbox on top of everything. Most importantly, we designed it with introverted and anxious students in mind, giving them an anonymous Q&A channel that most tools simply ignore.

What we learned

We realized quickly that UX matters just as much as the AI behind it, powerful models feel useless if the interface is confusing, and the Netflix-style metaphor made the whole experience click instantly for people. We also had to practice strict scope control: we deliberately cut real video generation, full PPT export, and multi-course analytics so we could ship a clean, reliable demo that actually works end to end. Along the way, we saw how good prompts are basically half the product; tiny phrasing tweaks like “Return JSON ONLY in this format…” dramatically changed how often the AI behaved itself. And because this is an education tool, we had to respect real-world constraints: we focused on using only content that students or instructors are actually allowed to upload, and avoided any design that encourages sneaky lecture recordings. All of that shaped how we framed, built, and positioned BadgerFlix.

What's next for BadgerFlix

In future, we would be using Trivi, which is a CNCF tool that detects known vulnerabilities in the image/video before deployment. We would love to add richer study tools like one-click flashcards for each episode, auto-generated quizzes, and downloadable “episode summary” PDFs or slide decks. We also want to integrate with real learning platforms like Canvas or Moodle, so instructors can send lectures directly to BadgerFlix and embed the generated episodes back into their LMS. Longer term, we see personalized learning paths: tracking what a student has watched, recommending the next episode or topic they’re weak on, and adapting quizzes based on their performance. On the instructor side, we’d like to ship better tools and analytics, things like seeing which episodes cause the most confusion, and automatic clustering of similar student questions so they can answer a whole “theme” at once instead of repeating themselves.