sartor87/ai-engineer-interview-concept

This project is a study tool built to help someone prepare for an AI Engineer interview that covers eight core topics. It combines a browsable set of interview material with a chat panel where you can practice answering questions and get feedback from an AI model. The front end is a React application built with Vite, and it talks to a backend written in .NET 8 that runs as an Azure Function. That backend does two main jobs: it checks each message with a guardrail service called Stihia before anything reaches the AI, and it forwards approved requests to an AI model called NVIDIA Nemotron Mini, streaming the reply back to the browser as it is generated. The whole thing is meant to run on Azure Static Web Apps, using the free tier, with the infrastructure defined in Terraform files so it can be recreated from code rather than clicked together by hand. Beyond the app itself, the repository documents its own architecture in detail. There is a set of C4 style diagrams describing the system at different zoom levels, plus a folder of short decision records explaining choices like why Azure was picked, why the backend uses .NET 8, and why the interview practice is split across three separate AI agents, an interviewer, an examiner, and a follow up agent. A GitHub Pages workflow automatically turns these diagrams and decision records into a small static website whenever the architecture folder changes, so the documentation stays visible without extra manual work. Running it locally needs Node.js 20 or newer, the .NET 8 SDK, and the Azure Functions command line tools, since you start the frontend and the backend function as two separate processes and let a proxy connect them. A Stihia API key is required to use the guardrail checks, though the project mentions a setting to turn guardrails off for local testing. The project also ships a fair amount of tooling for the code itself, including custom Claude Code configuration files, lint checks, and a pre-commit script covering code quality and architecture consistency, which suggests it is meant as much as a demonstration of a well organized engineering setup as it is an interview study aid. The full README is longer than what was shown.