FriendPrep AI: An AI Interview Partner for a Friend

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

What I Built

I built FriendPrep AI, an interview practice app for a friend preparing for job interviews. Instead of practicing with the same generic question list, they can enter their target role, skills, experience, resume, and a job description to get a more tailored practice session.

The app generates interview questions, evaluates answers with scores and feedback, can ask follow-up questions, and creates a final report with areas to work on. I also worked on making the feedback honest: answers that don’t address the question shouldn’t be praised as strong.

Demo

Live demo: https://friendprep-ai.onrender.com/

The app is deployed on Render.

Code

GitHub repository: FriendPrep AI

How I Built It

FriendPrep AI is built with Django and PostgreSQL. Its AI layer uses the open-weight Qwen3-4B-Instruct-2507 model through Hugging Face Inference Providers.

The app keeps provider calls behind an AI service layer, so question generation, answer evaluation, follow-ups, and the final report are handled separately from the Django views. That also makes it easier to change the model or inference provider later.

Why Does Open Innovation Matter?

Interview answers and resumes can be personal, so it matters that the model isn’t a black box tied permanently to one provider. Using an open-weight model gives the project room to change models, inspect available model options, or eventually self-host inference.

For this version, inference is hosted through Hugging Face, so candidate information is still sent to that provider. The open model gives the project a path toward more control later; I’m not claiming the current hosted setup keeps that data entirely local.

Prize Categories

• Best Use of Render — for hosting the app and its interview-practice experience on Render.

• Best Use of GitHub Copilot — for the coding-agent assistance used throughout the development process.