VivaBot: The AI External Examiner

opening page
home page where bot actually asks questions and checks our answers
sample final report generated
pdf uploading option

Inspiration

Oral vivas scare students, and external examiners have limited time. I wanted an AI examiner that students could practice with anytime before real exams.

What it does

VivaBot simulates an external examiner. A student selects a topic or uploads their notes/PDF, and the bot asks viva-style questions one by one, evaluates each answer from 0–100, gives feedback, and finally generates a report card summarizing strong and weak areas.

How we built it

The backend is built with Django. It calls Google Gemini models through the Python client to generate questions, evaluate answers, and create the final report. The frontend is a simple responsive web UI with a landing page, chat-style interface, and loading animations. The app is deployed on Render as a live web service.

Challenges we ran into

Managing API rate limits and key rotation for Gemini.
Cleaning and parsing model output reliably for the scoring and report card.
Balancing hackathon work with university semester exams and limited time.

Accomplishments that we’re proud of

End‑to‑end viva flow from topic/PDF input to final report card.
A live deployed web app that judges can use directly in the browser.
Turning exam anxiety into a guided practice experience for students.

What we learned

I learned how to integrate LLMs into a production‑style Django app, handle prompt design for evaluation, and deploy a full‑stack AI project on a PaaS (Render).