Maaahr

Interview advice is easy to read and harder to put into practice.

Maaahr started as a text based AI Engineering capstone. I wanted to make interview practice easier to do, especially when finding someone to practise with was difficult. During a hackathon, I rebuilt the interview around the OpenAI Realtime API so it could listen and speak. I kept building around that conversation, adding a place to choose a role, practise again and see what needed work.

The practice app lets someone choose a position from a library or describe their own. The interview uses that role's description, experience level and focus areas to guide the questions. They can answer through text or voice, then review a summary, strengths and feedback on each answer. I kept previous attempts under the same position so candidates could compare sessions and see how their answers changed.

1. 01

   Arabic and English throughout

   I built Arabic and English into the questions, prompts, interfaces, reports and emails. The app supports right-to-left and left-to-right layouts, helping job seekers prepare for interviews in either language, as the market requires.

2. 02

   A clear start

   The interview app introduces the session, explains the available answer formats and timing, then brings the candidate into the conversation.

I also built a workflow for hiring teams around the same interview. A company creates a position with its questions, focus areas and time limits, then shares a link with candidates. The dashboard brings the responses together, with a summary and feedback for each candidate. This gave the product two different entry points while keeping the interview and its results connected.

Voice practice should feel like a real exchange.

I shaped the interview so candidates could focus on their answers, not the screen. Clear cues show when the interviewer is speaking and when it is their turn. A pause to think still feels like part of the conversation, and the exchange keeps its rhythm from the first question to the last.

1. 01

   The same path to feedback

   I saved the interviewer's transcript as a question and the candidate's transcript as an answer. Text and voice sessions could then use the same feedback flow, result page and email.

Between December 2024 and August 2025, I took Maaahr from a text capstone to four connected apps with voice interviews. The screenshots show the interfaces with sample data and a demo session. The lesson I took from it was how much voice depends on timing. Connecting an audio model was only part of making the interview feel natural.

1. 01

   What I'd work on next

   I'd add recording and review so people could hear an answer beside its feedback. I'd also explore different interviewer styles and compare the scoring with real hiring decisions. The score is a part of the product I would still want to calibrate.