Gemma Viva Drill: An Open-Source Viva and Concept Examiner Built for My Friend

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend.

What I Built

Preparing for college viva exams, technical interviews, and core computer science oral assessments is usually stressful. Reading textbooks or scrolling through passive flashcards creates an illusion of competence, but it fails to simulate the pressure of spontaneous oral recall. Classmates are not always available to conduct repetitive drill sessions, and solo mock trials feel unnatural.

I built Gemma Viva Drill for my friend and classmate who is currently preparing for technical vivas and campus interview rounds. He frequently understood core engineering concepts on paper, but froze or skipped key steps when questioned directly in viva sessions.

Gemma Viva Drill functions as an interactive, rapid-fire oral examiner. It:

1. Poses focused, one-at-a-time conceptual questions in Computer Networks, Operating Systems, and Data Structures.
2. Evaluates the student answer immediately with clear verdicts: Correct, Partially Correct, or Needs Work.
3. Provides a concise, two-sentence explanation to correct misconceptions.
4. Immediately serves the next logical, probing follow-up question to keep the pace brisk.

Demo

Live Deployed Application: https://gemma-viva-coach.onrender.com

GitHub Repository: https://github.com/ankiitmishraa-tech/gemma-viva-coach

How I Built It

1. Open-Weight AI Engine: Powered by Google Gemma 2 (9B-IT), guided by structured system instructions to ensure rapid conversational viva evaluation without long-winded lecturing.
2. Web Interface: Developed with Python and Streamlit, maintaining clear chat state and immediate question rendering.
3. Production Hosting: Deployed continuously as an active web service on Render.

Why Open Innovation Matters

Closed-source, commercial AI APIs enforce strict usage metering, continuous token expenses, and privacy compromises during intensive, multi-hour study marathons.

Adopting an open-weight foundation model like Gemma changes the equation:

1. Zero Cost Lock-in: Students can run endless mock drills without dreading recurring pay-per-token API bills.
2. Offline Portability: Because Gemma is open-weight, this pipeline can easily be transitioned to fully offline inference on a laptop using local engines like Ollama, making it reliable in connectivity-constrained environments.
3. Pedagogical Customization: Open weights let developers calibrate system behavior, domain boundaries, and strictness precisely, without proprietary safety filters derailing technical discussions.

The Handoff

I handed the live link over to my friend today to run an intensive drill on Computer Networks, specifically covering TCP handshakes and subnetting.

Here is his feedback:
"Normally when I read notes, I convince myself that I know the topic. But when an examiner actually asks a question on the spot, I hesitate. Having this tool point out my mistake in two sentences and immediately hit me with the next question felt like a realistic viva without the anxiety."

Prize Categories Targeted

1. Best Use of Gemma ($200)
2. Best Use of Render ($200)