BUILD FOR FRIEND CHALLENGE

Rohit Kumar singh — Mon, 05 Oct 2026 06:38:21 +0000

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

What I Built

I built SpeakBuddy, a private AI communication practice partner for a friend who gets nervous and loses confidence while speaking English during interviews, presentations, and conversations.

My friend often knows what they want to say but struggles to express their thoughts clearly and confidently. SpeakBuddy provides a safe space to practice without the pressure of a real interview or presentation.

The user can choose between Interview, Presentation, and Conversation modes. SpeakBuddy asks a question, analyzes the user's answer using AI, provides scores for Clarity, Grammar, Vocabulary, and Confidence, gives improvement suggestions, and generates a follow-up question for continued practice.

Demo

Live Demo: https://speakbuddy-dnra.onrender.com/

Video Demo: https://drive.google.com/file/d/1e8jhW-B7tqM9UsAffvUzhhPyfrWRvC8Z/view?usp=sharing

Code

GitHub Repository: https://github.com/rohit060321/SpeakBuddy.git

How I Built It

The core of SpeakBuddy is built around Gemma 3 1B, an open-weight AI model running locally through Ollama.

The application is built using Python and FastAPI for the backend, HTML/CSS/JavaScript for the frontend, MongoDB Atlas for storing practice sessions, ElevenLabs for voice feedback, and Render for deployment.

The basic flow is:

Question → User Answer → Gemma Analysis → Scores → Suggestions → Follow-up Question

Gemma analyzes the user's response and returns structured feedback containing scores for clarity, grammar, vocabulary, and confidence, along with improvement suggestions and a follow-up question.

Why Does Open Innovation Matter?

Open innovation was important for SpeakBuddy because communication practice can involve personal answers, experiences, and weaknesses.

Using an open-weight model like Gemma gave me more control over how the AI works instead of depending completely on a closed AI API.

Running Gemma locally through Ollama also made experimentation easier. I could modify prompts, change the feedback structure, test different behaviors, and build the application around the model rather than treating the AI as a black box.

For this project, open innovation made it possible to build an AI tool that is more customizable, transparent, and privacy-conscious.

My Agent Session

I used AI-assisted development while building SpeakBuddy to help with implementation, debugging, and improving the project.

Prize Categories

Gemma / Hugging Face — Gemma 3 1B is used as the core AI model through Ollama.
MongoDB Atlas — used for storing practice sessions and history.
ElevenLabs — used for voice feedback.
Render — used to deploy the application.

hellloooo everone

Rohit Kumar singh — Mon, 05 Oct 2026 01:29:50 +0000

DEV Community: Rohit Kumar singh