
Interview practice is hard to come by. Mock interviews with real people are expensive or difficult to schedule, and typing answers into a chatbot doesn't feel like a real interview. Prep.AI solves this by letting candidates practice the way interviews actually happen: by speaking. How it works: the user signs up, chooses an interview, and receives AI-generated questions. They answer out loud using their microphone. The recorded audio is sent to our FastAPI backend, where AssemblyAI transcribes the spoken answer into text. That transcript is then evaluated by an LLM (via Groq), which returns a score and specific feedback on the answer's content and structure. Key features: - Voice answers transcribed with AssemblyAI speech-to-text - AI-generated interview questions - Per-answer scoring and actionable feedback - Analytics dashboard to track progress across sessions - Secure JWT authentication and PostgreSQL storage - A pet reward system to keep users motivated Tech stack: React (Vite) frontend, FastAPI backend deployed on Render, PostgreSQL on Neon, AssemblyAI for transcription, and Groq for question generation and evaluation. Prep.AI turns solo practice into a realistic, spoken interview experience with instant, honest feedback.
30 Sep 2026