Think Out Loud!

AI-powered interview simulator to practice oral exams and job interviews.

Comment

Inspiration

As a student pilot, we are required to take an oral exam alongside our practical flying exam to demonstrate our proficiency in aeronautical knowledge. Studying flashcards and reading books is always helpful, but interactive learning is what really helps you get better at formulating cohesive thoughts that make sense to whoever is interviewing you. If I want to pass my oral exam this December, I need to be able to say my answers with confidence.

What it does

Think Out Loud! is an interactive AI-powered oral exam simulator. You enter a prompt—like “Private Pilot Checkride” or “Software Engineering Interview”—and the app generates a set of questions. It will read them out loud to you, just like an examiner would. You respond with your voice, and the system listens, evaluates, and gives you feedback. No typing, no flashcards—just natural conversation to help you study and prepare more effectively.

How we built it

The frontend is built with Next.js and Tailwind CSS. The backend is written in Go, which handles audio and AI transcription, along with generating the exam questions:

  • Deepgram provides fast and accurate speech-to-text transcription of the user’s spoken answers.
  • Google Gemini powers the LLM evaluation, generating follow-ups and feedback to simulate a real oral examiner.

Challenges we ran into

Because I've had teammates bail on my on multiple occasions, I did not end up having an idea for my project until about 10pm. I was always hesitant and stubborn when it came to using AI and LLMs, especially considering I've had zero experience with both so this project definitely pushed me out of my comfort zone.

Accomplishments that we're proud of

  • Building a voice-only learning flow from scratch that feels smooth and natural.
  • Making the system field-agnostic—from aviation checkrides to tech interviews.
  • Successfully blending Go, Deepgram, and Gemini into a unified backend pipeline.

What we learned

When I thought of my idea initially, I thought it wouldn't be doable because I didn't have a clue what I was doing when it came to AI. But after some hours of research and playing around with APIs like OpenAI, Gemini, Deepgram, I've finally come to accept to using AI in my projects.

What's next for Think Out Loud!

I'm actually super impressed with how everything came out. My website works VERY well if you fine-tune your prompt good enough. I will be pursuing this project in my pastime after MHacks to turn it into a real startup - and so I can pass my Private Pilot checkride!

Built With

Share this project:

Updates