I Built an Offline AI Interview Coach for My Friend's Job Hunt

This submission was part of the Hacktoberfest Weekend Challenge, where the goal was to build something useful for a friend. The author created an offline AI interview coach to help a friend preparing for software engineering interviews. The coach provides realistic mock interviews, asks technical and behavioral questions, adapts follow-up questions based on the candidate's answers, explains strengths and weaknesses, suggests better answering structures, generates additional questions, and provides a final interview report.

The AI used in the project is Gemma, an open-weight model, which allowed the developer to run, inspect, modify, and experiment with the AI rather than sending conversations to a closed API. The coach separates the interview workflow from the model itself, enabling experimentation with different models, prompts, context, and evaluation logic independently.

The key design decision was that the AI's purpose is not to do the interview preparation for the user, but to facilitate practice. The AI doesn't reveal the correct answer immediately; instead, it tries to act like a real interviewer by asking the user to explain their approach. Only after the interview does it provide feedback and suggestions for improvement.

The author emphasizes that the main lesson learned was that building an AI application is about designing the interaction around the user's problem rather than just producing impressive text. The open innovation approach allowed the project to focus on control, privacy, and experimentation, making it more useful for the specific user case.