Perception AI
AI-powered investor sentiment analysis platform with Gemini Live API voice interviews and Firebase Cloud Functions.
What I did
Architected full-stack AI-powered investor sentiment analysis platform using Next.js 16, TypeScript, Firebase Cloud Functions, Firestore, and Google Gemini
Built real-time conversational voice interview system using Gemini Live API WebSocket streaming with bi-directional audio, implementing 6-phase state machine
Engineered audio processing pipeline with Web Audio API handling 16kHz PCM recording, 24kHz playback resampling, and echo feedback prevention
Developed AI chat interview system with Gemini conversational intelligence, exponential backoff retry logic, and sophisticated answer extraction
Implemented intelligent question generation with Gemini structured output and semantic Likert scale detection across 7 scale types
Created 20 Firebase Cloud Functions including AI question generation, bulk email invitations via Resend API, and Google Cloud TTS/STT integration
Built multi-tenant organization system with custom AI prompt injection enabling personalized AI interviewer behavior per organization
Developed survey analytics dashboard with sentiment gauge visualization, Likert distribution charts, and AI-generated theme cards