01
Real-Time Voice Activity & ASR
Continuous Voice Activity Detection (VAD) coupled with fast Automatic Speech Recognition (ASR) captures spoken answers naturally with minimal latency and high transcription fidelity.
Coming soonIn active development
Vakya
Master interviews, regardless of your English background.
An AI-powered spoken interview coach designed to bridge the gap between candidate capability and interview communication. Practice answering questions out loud with real-time voice activity detection (VAD), automatic speech recognition (ASR), conversational text-to-speech (TTS), and instant critique on grammar and clarity.
Project Status
Private Development & Testing
Vakya is currently being tested on staging environments with select users to refine real-time voice latency across VAD and ASR pipelines, conversational TTS pacing, and evaluation rubrics. Public access will launch soon.
How it works
02
Conversational TTS & AI Critique
Natural Text-to-Speech (TTS) models conduct the mock interview out loud, while evaluation models analyze responses to help reshape casual phrases into corporate-ready answers.
03
Full-Duplex WebSockets & Audio Streaming
Low-latency bidirectional event and audio communication handled in real-time by Laravel Reverb, with candidate history, scores, and analytics persisted in PostgreSQL.
Architecture & Technologies
Backend & Streaming
Laravel · WebSockets · Reverb · PostgreSQL
Audio & AI Models
ASR model · TTS model · VAD model
Frontend
React · TypeScript · Tailwind CSS
Interested?
If you want to try an early build, collaborate, or discuss real-time speech and interview tech, send an email to hi@debjit.in.