I Built an AI Interview Coach for My Friend — Powered by Open-Weight AI

Roushan kumar jha — Mon, 05 Oct 2026 17:46:43 +0000

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

What I Built

I built Interview AI, a personalized interview-preparation assistant for a friend who was actively preparing for software engineering interviews.

The problem was simple: generic interview-preparation tools can generate questions, but they don't necessarily understand the specific candidate, their projects, their experience, or the role they're applying for.

Interview AI turns that preparation into a personalized experience.

A candidate can provide their background and the role they're preparing for, and the application generates relevant interview questions and helps them practice answers based on their actual experience.

The goal wasn't to build another general-purpose AI chatbot. I wanted to build something small that my friend could actually use before an interview.

Demo

Live application: https://interview-ai-azure-eight.vercel.app/

The application is designed around a simple loop:

Candidate context → Interview question → Candidate answer → AI feedback → Better preparation

Code

GitHub: Repository

The project is built as a web application with a focus on keeping the interview workflow simple and usable rather than turning it into a generic AI chat interface.

How I Built It

The application is built with Next.js, TypeScript, and an open-weight AI model at its core.

The AI is responsible for understanding the candidate's context and generating interview interactions rather than simply returning predefined questions.

The basic architecture is:

Candidate
    ↓
Next.js Application
    ↓
Interview Context
    ↓
Open-weight AI Model
    ↓
Personalized Interview Question
    ↓
Candidate Answer
    ↓
AI Evaluation & Feedback

I intentionally kept the AI layer replaceable. The application isn't fundamentally tied to a single closed AI provider, which means the underlying model can be changed, experimented with, or run through different inference setups as the project evolves.

The most important part of the implementation is that the open AI model isn't just an optional feature. It is the component responsible for creating the personalized interview experience.

Why Does Open Innovation Matter?

For an interview-preparation tool, the candidate's information can be highly personal: their resume, projects, previous experience, weaknesses, and the roles they're targeting.

Using an open model gives me significantly more control over that AI layer.

Instead of treating the model as a black-box API that I cannot inspect or replace, I can choose the model, change the inference setup, experiment with prompts and behavior, and eventually run the system locally if the use case requires stronger privacy.

That matters especially for a tool built around someone's personal career information.

Open innovation also changes the economics of experimentation. I can prototype, evaluate different models, and potentially run inference without being permanently locked into a proprietary API or its pricing and availability.

For this project, open AI means ownership and flexibility over the part of the product that actually makes it useful.

My Agent Session

Prize Categories

Gemma

I am entering the Best Use of Gemma category because the project uses Google's open-weight Gemma model as the AI foundation for the interview experience.

The model is not being used simply as an add-on chatbot. It is responsible for generating and evaluating personalized interview interactions based on the candidate's context.

Built for one person, but designed around a problem many candidates face:

How do I practice for my interview rather than just practice interviews in general?

DEV Community: Roushan kumar jha