Smart Interview Companion

Smart Interview Companion simulates real technical interviews using conversational AI, generating questions and voice feedback to help you practice, improve, and gain confidence.

Comment

Inspiration

As technical interviews become more rigorous and competitive, many candidates, especially early-career developers struggle to find structured, competitive and realistic practice environments. I wanted to build a solution that not only mimics real-world interviews but also provides dynamic, voice-enabled interaction powered by the latest in Generative AI tools. The goal was to bridge the gap between mock interviews and actual interview pressure.

What it does

Smart Interview Companion is an AI-powered mock interview platform that simulates realistic technical interviews. It

  • Generates mock and actual interviews questions using Perplexity’s sonar-pro model. The interview can be configured based on
    • Level -> Senior, Mid, Entry
    • Domain -> Software Engineering, Data Science etc
    • Duration
    • Interview Type -> Technical, HR, Case Study etc
  • Enables natural voice interaction with ElevenLabs TTS integration and gives a natural interview space
  • Provides structured sessions with introductions, questions, and feedback.
  • Supports both text and voice-based replies from the candidate
  • It tailors interviews by region and role and asks relevant questions based on the same.
  • It can follow up with valid follow up questions from the answer given.
  • It is Mobile first designed web app giving a very good experience on Mobile.

How we built it

The Tech Stack chosen are Frontend: Built using Bolt.new for fast, responsive UI components. Backend: Developed with FastAPI, handling user sessions, prompt generation, and AI model orchestration. Database: Supabase for fast database integration and user auth and sessions AI Models: Perplexity Sonar-Pro: For generating realistic technical questions and feedback ElevenLabs TTS: For converting interviewer prompts into lifelike voice responses.

The UI is completely built with bolt.new. Initial build, improvements and bug fixes done purely with help of the platform and prompting. Backend is a quick development done with FastAPI with Sonar-Pro integration

Ops FastAPI -> Deployed on Render Free tier Client [ Bolt.new ] -> Netlify native deployment on free tier Database -> Supabase

Challenges we ran into

Here are some of the Major challenges that I ran into in the order

  1. Coming up with the right prompts and making it dynamic and robust based on the configuration the user chooses. With each config change the prompt modifies and assembles to give the right set of question and experience for the user.
  2. Iterative building of UI with bolt.new with prompts. Even if the UI build was done as Breeze with zero coding, there was some struggle in getting minute changes in. Sometimes the changes made will break the app completely and had to revert them through github and prompt again multiple times. One major part where it got stuck was when the Interview Page failed to load due to some enhancements done on Theme.
  3. Proper voice synchronization in the app with Elevenlabs. The current integration is not production grade, but to bring it to this level there was some difficulty.

Accomplishments that we're proud of

  1. Built a fully functional, AI-driven interview preparation tool MVP within the hackathon timeframe. That I feel like a huge accomplishment.
  2. Successfully integrated multimodal AI (text + voice) and got it working
  3. Making structured interview sessions that feel human-like and adaptive.
  4. Last but not least, was able to get over challenges without workarounds.

What we learned

  1. Building a Real-time Full Stack AI powered Mobile First designed web app. Not only targeting for hackathon but to take it further as a Micro Startup to open this up to all and make Preparing for Interview sessions cooler and easier.
  2. How to use AI powered tools like bolt.new to kick start the startup journey and make a quick MVP of the product in lesser days.
  3. Learnt advanced prompt engineering techniques for maintaining contextual conversation for the interview process. With that also to generate various versions of the same prompt dynamically based on user interaction
  4. How to orchestrate multiple AI services (LLMs + TTS) into a single smooth user flow.
  5. Managing stateful conversations in stateless HTTP environments using modern backend frameworks like FastAPI.
  6. How to manage time from the Real time job to build a Solo Micro Startup.

What's next for Smart Interview Companion

There is a lot I need to do to convert Smart Interview Companion as a Micro startup and release it soon. The roadmap will be as follows In next 1 month [ Make the App Better ]

  • Integrate secured API calls with JWT Authentication in FastAPI.
  • Integrate social authentication.
  • Add email verification on signup.
  • Integrate payment method for Pay and Use in the app.
  • Better prompt response parsing. Some responses are not parsed properly and formatted.
  • Make the interview interactions more fluid. Move away from HTTP API calls for the same and try using websocket / gRPC
  • Improve the AI feedback at the end as voice interactions. And more insights to better prepare for the interview
  • Make the prompt generation multi model instead of relying on sonar-pro alone. Pick the best model for questions, feedbacks, research etc and have back up models also for each based on availability.

Short term Main Features plan

  • Add Interview preparation based on Job Description. Copy paste your desired job title and description. Tune the config and get ready for the session. [ This was done before in MVP but had to revert due to some UI issues and getting late for submission ]
  • Add FAANG interview preparations
  • Add Interview Gamifications and leaderboard. Compete with other users and expose your interview strengths for others to see and prepare
  • Get your favorite company insights and interview process for better preparation

Long term Features [ If Short term works well ]

  • Sync your LinkedIn profile and get pointers to improve your profile based on the current industry insights
  • Integrate with Job Search. Favorite your dream job and start preparing for it immediately. Get a detailed road map for achieving your goal.
  • Form a community and Connect with others, learn to crack interview together

Built With

  • bolt.new
  • chatgpt
  • claude
  • elevenlabs
  • fastapi
  • github
  • netlify
  • perplexity
  • python
  • react
  • render
  • sonar-pro
  • supabase
  • tailwindcss
Share this project:

Updates