VivaShield AI: I built the open-source viva defense tool that predicted 4 out of 5 actual examiner trap questions

Description and use

VivaShield AI is a local, open‑source tool that uses Gemma 2B to predict real examiner trap questions for capstone defenses, providing defenses and a readiness score while keeping code private and cost‑free.

Problem

Engineering students face unpredictable examiner trap questions during final-year viva defenses, risking failure and privacy concerns.

Outcome

The tool accurately predicted a real examiner question, providing a bulletproof defense that matched the actual query. It operates entirely offline, ensuring code privacy and zero marginal cost.

Workflow

• Clone the repo
• Navigate to the project directory
• Pull the Gemma 2B model with Ollama
• Run the provided script to start the local server
• Paste the project README, tech stack, and a key code snippet into the UI
• Receive generated trap questions, defenses, Q&A, scenarios, cheatsheet, and readiness score
• Print or export the 1‑page cheatsheet for pre‑viva review

Evidence

The student who tested this was asked exactly this question by their external examiner.

Source

Runs 100% locally using Google Gemma 2 via Ollama.

Source