Bojon

Got no Job? Practice for your interview with Bojon

Comment 5

Inspiration: No Job, No Problem

The job search can be brutal—especially for recent grads or folks re-entering the workforce. Interviews are often the bottleneck, and yet there's no fun or effective way to actually practice them. We wanted to change that. Inspired by the need for real-time, low-pressure interview practice, we built Bojon: a competitive, 1v1 web-based interview game designed to gamify job prep and help users level up through meaningful feedback.

What It Does

Bojon connects two users in a live 1v1 face-off where each player gets the same technical interview question and has 20 seconds to respond on video. Once the timer ends, their video response is automatically analyzed and scored using Twelve Labs' multimodal video understanding API. The best part? No judges, no waiting, no awkward silences—just pure, fast-paced interview prep with instant feedback.

Features include:

  • Matchmaking lobby (built without Express)
  • Live question generation using Gemini Pro
  • Video recording and upload directly in-browser
  • AI-powered feedback using Twelve Labs
  • ELO-based ranking system to gamify improvement

How We Built It

We built Bojon in 24 hours using:

  • Vanilla HTML/CSS/JS for the frontend (no React)
  • Node.js HTTP server (no Express) hosted on a Raspberry Pi
  • MongoDB for user data and ELO storage
  • Gemini Pro API for real-time interview question generation
  • Twelve Labs API for video understanding and performance scoring
  • Ribbon AI API tested for audio integration (experimental)

We intentionally avoided frameworks and kept things minimal to focus on speed, performance, and creativity.

Challenges We Ran Into

  • Building a full stack server on vanilla Node.js without Express required careful routing logic and async handling.
  • Managing real-time matchmaking and race conditions between users was tricky.
  • Video capture and encoding in-browser required handling multiple browser quirks.
  • Integrating and streaming video to Twelve Labs while managing local file conversions on a Raspberry Pi added extra complexity.
  • Ensuring reliable UX in a tight loop (match → record → upload → feedback) without bugs or lag.

Accomplishments That We're Proud Of

  • Fully working multiplayer interview game in under 24 hours.
  • Seamless AI integration with Gemini and Twelve Labs for dynamic content and feedback.
  • Beautiful, cyberpunk-inspired UI without a single UI framework.
  • Full deployment on a Raspberry Pi with zero external dependencies like Express or React.

What We Learned

  • How to build fast, clean backends using only core Node.js
  • How to stream and process video on-device while integrating AI services
  • How to manage game logic, player state, and API concurrency in a server-light architecture
  • The power of Gemini and Twelve Labs in building meaningful, interactive experiences with AI

What's Next for Bojon

  • Add real-time multiplayer chat and voice
  • Implement leaderboard and match history
  • Allow users to write and rate each other's answers
  • Support behavioral and system design interview tracks
  • Enable mentor mode with human reviewers
  • Deploy to scalable cloud infrastructure

Bojon was built to make interview prep less painful and a lot more fun. We're just getting started.

Built With

Share this project:

Updates