Inspiration Job interviews can significantly affect a person’s career, yet many candidates prepare alone with static question lists, generic chatbots, or expensive mock-interview sessions. We wanted to create a more realistic and accessible way to practice—one that reacts to what the candidate actually says and provides clear guidance on how to improve.
InterviewAI was inspired by the gap facing students, fresh graduates, job seekers, and non-native English speakers across Egypt and the MENA region: they need practical interview training that is available anytime, personalized to their target role, and comfortable to use in either English or Arabic.
What it does InterviewAI is an AI-powered interview simulator that helps users practice HR, behavioral, and technical interviews.
Users can choose a target role, experience level, interview type, language, and interviewer style. The platform conducts a realistic interview one question at a time, accepts voice or typed answers, produces a live transcript, and generates adaptive follow-up questions based on the user’s response.
After each answer, InterviewAI provides structured feedback covering:
Clarity, relevance, structure, evidence, and confidence.
Strengths, weaknesses, and practical suggestions.
Speaking pace and filler-word analysis.
STAR-method feedback for behavioral answers.
A rewritten model answer to demonstrate a stronger response.
The platform also provides a final performance report with answer-by-answer feedback, skill analysis, and a personalized practice plan. Users can optionally provide their resume and a job description so that questions are tailored to their real experience and the role they are targeting.
How we built it We built InterviewAI as a full-stack web application using:
Next.js 15, React, TypeScript, and Tailwind CSS for the frontend and user experience.
Vercel for deployment, hosting, and CI/CD.
Clerk for authentication, user accounts, and protected dashboard routes.
PostgreSQL on Neon to store interview sessions, questions, answers, evaluations, and user progress.
Prisma ORM to connect the application to the database.
Google Gemini API to generate adaptive questions, evaluate responses, create follow-up questions, and produce final reports.
unpdf and mammoth for extracting text from uploaded PDF and DOCX resumes.
@vladmandic/face-api with TensorFlow.js for privacy-first body-language analysis running in the browser, so camera frames do not need to leave the user’s device.
Challenges we ran into One major challenge was making the AI feedback useful instead of vague. A simple score is not enough for interview preparation, so we designed structured evaluation criteria and focused on giving feedback that tells users exactly what to improve.
Another challenge was building adaptive interview behavior. The interviewer needs to ask relevant follow-up questions based on the candidate’s previous answer rather than behave like a static question generator.
We also had to consider privacy carefully because the product can use a microphone, camera, resume, and job description. Our approach is privacy-first: recording is opt-in, and body-language processing is designed to happen locally in the browser rather than uploading camera frames.
Finally, supporting both English and Egyptian-ready Arabic creates additional complexity in transcription, filler-word detection, prompt design, and producing feedback that feels natural in both languages.
Accomplishments that we're proud of Built a working AI interview practice experience rather than a static question bank.
Created adaptive questions and follow-ups that respond to the candidate’s answers.
Combined voice or typed answers, live transcription, AI evaluation, and performance reporting in one workflow.
Added practical feedback such as strengths, weaknesses, STAR-method analysis, and rewritten model answers.
Designed the platform around privacy, including opt-in recording and on-device body-language analysis.
Built a complete product foundation with authentication, protected user sessions, a database-backed interview history, and cloud deployment.
Defined a realistic business model with freemium access for individual users and future B2B licensing for universities, bootcamps, and career centers.
What we learned This project taught us that building an AI product is not only about connecting an LLM API. The real value comes from designing a complete user journey: asking the right question, interpreting the answer in context, presenting feedback clearly, and helping the user take the next improvement step.
We also learned the importance of prompt engineering and evaluation rubrics. Different interview types require different criteria: a behavioral answer should be evaluated differently from a technical or system-design answer.
Privacy was another important lesson. When a product handles sensitive content such as resumes, voice responses, and camera input, privacy should be part of the architecture from the start—not an afterthought.
What's next for InterviewAI Next, we plan to improve the platform through:
Stronger resume and job-description matching for more role-specific questions.
More interview modes, including frontend, backend, AI/data, HR, behavioral, and system-design interviews.
Deeper progress analytics showing trends across multiple practice sessions.
Downloadable AI-generated performance reports.
Improved Arabic support, including Egyptian dialect handling and Arabic-English technical language switching.
More accurate and user-controlled body-language insights.
Institution dashboards for universities, bootcamps, and career centers.
A freemium plan with advanced analytics, unlimited practice sessions, and priority AI access for premium users.
InterviewAI’s goal is to become an always-available, privacy-conscious interview coach that helps candidates practice with confidence before a real opportunity.
Built With
- n/a
Log in or sign up for Devpost to join the conversation.