What I Built
I built DevRehearsal, a 100% local, voice-enabled AI Oral Examiner & Technical Tutor designed to help students and developers practice project defenses, viva exams, and technical hackathon presentations in a safe, interactive environment.
The Story & Who I Built It For
As computer science students, my friend and I frequently build projects and participate in hackathons. However, whenever a major college viva, project evaluation, or hackathon defense approached, we found ourselves facing the exact same challenge: a lack of confidence in articulating what we built.
Even when our code was fully functional and our architecture was solid, sitting across from a tough examiner or hackathon judge made us second-guess ourselves. Practicing in front of a mirror or reviewing notes quietly didn't work—it couldn't simulate the pressure of being asked sharp, unpredictable questions on edge cases, tech stack choices, or system trade-offs. We needed a patient, intelligent practice partner that could challenge us, point out our blind spots, and teach us how to structure clear, confident technical answers.
That is why I built DevRehearsal for my friend and fellow student developers facing the exact same anxiety.
What It Does
Generates Topic-Specific Defense Questions: Acts as a strict technical examiner asking questions tailored to selected topics (e.g., Computer Vision, Python & Data Structures, React & Node, System Design).
Evaluates Answers with Step-by-Step Feedback: Scores responses out of 10, highlights key strengths, pinpoints technical errors, and outlines an ideal step-by-step answer.
Interactive Voice Capabilities: Uses native Web Speech synthesis (TTS) to speak questions out loud and Web Speech recognition (STT) for hands-free voice responses.
Local First with Simulation Fallback: Connects directly to local Ollama endpoints for 100% private inference, while offering a seamless fallback simulation mode.
Demo
Live Application
- Deployed Web App: https://vikas-verma1.github.io/devRehersal/
Application Interface
Code
This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
What I Built
Frontend Dashboard Preview
DevRehearsal is a 100% local, voice-enabled AI Oral Examiner & Technical Tutor designed to help students and developers rehearse project defenses, viva exams, and technical hackathon presentations.
I built this specifically for my classmate and friend who was feeling anxious about answering unpredictable technical questions during upcoming college project viva exams and hackathon presentations.
The Problem It Solves:
- Presentation Anxiety: Practicing alone is ineffective because you can't anticipate tough questions.
- Lack of Constructive Feedback: Standard practice doesn't pinpoint why an answer was inadequate or show how to structure an ideal technical response.
- Privacy & Cost Concerns: Sharing proprietary project code or early hackathon ideas with commercial cloud APIs is risky, and API costs add up during repeated rehearsal loops.
Video Demo
Watch the DevRehearsal Demo Video
Requirements & Prerequisites
Before running…
Repository Link
- GitHub Repository: https://github.com/Vikas-Verma1/devRehersal.git
How I Built It
DevRehearsal is built using open-source AI models and modern web technologies:
-
Local Inference Engine: Connects directly to Ollama running on
http://localhost:11434. - Open-Weight Models: Designed to run with open-weight models such as Llama 3.2, Gemma 2, Qwen 2.5, and Mistral.
- Frontend Stack: Single-file HTML5 interface styled with Tailwind CSS, integrated with native browser Web Speech API for speech synthesis (TTS) and speech recognition (STT).
- Fallback Simulation Engine: Includes an automated simulation engine so users can explore the full UI/UX flow even when Ollama is offline.
Why Does Open Innovation Matter?
Open innovation was crucial to making DevRehearsal practical and accessible:
- Complete Code & Idea Privacy: Project details, early-stage hackathon code, and personal project architectures stay entirely on the user's local machine without being sent to commercial cloud servers.
- Zero Ingress/Egress Cost: Students can run unlimited practice sessions and mock interviews without worrying about token limits or API bills.
- 100% Offline Capability: Works completely air-gapped in college computer labs, hackathon venues, or areas with poor internet connectivity.
- Model Flexibility: Users can freely swap open-weight models depending on their hardware (e.g., fast 3B models on lightweight laptops, or 7B/14B models for deep technical evaluations).
Prize Categories
- Build for a Friend: Built specifically for a classmate and friend to help overcome presentation anxiety and build confidence for technical viva exams and hackathon defenses.
- Open-Source AI Challenge: Powered by local inference using open-weight models (Llama 3.2 via Ollama) as the core engine.
Top comments (1)
this is the my first post