The Problem We Found:
Candidates memorized interview answers on paper, but froze or rambled for 3 minutes without stating clear metrics when speaking out loud.
How We Solved It in Code:
Built Web Speech API real-time transcript streaming paired with GPT-4o mini STAR color-coding. Spoken text is highlighted live: Situation (15%), Task (10%), Action (50%), Result (25%).
Technical Stack:
- Web Speech API + Tailwind HSL tokens + JSON-schema prompt parsing.
- Result: Reduces answer rambling by 40% in candidate practice sessions.
Check out the live build: https://interview-prep.io
#buildinpublic