In-House Innovation // Proof of Work
Dincharya — Intelligent Routine OS
We don’t just talk about full-stack engineering; we build, deploy, and scale our own software products. Dincharya showcases our end-to-end craft in AI streaming, voice processing, and modern cloud architecture.
Technology Blueprint
Next.js 15SupabaseGroq WhisperLlama 3.3PostgreSQLTailwind CSSZod
Audio AI Pipeline
Groq Whisper Voice Engine
Records voice reflections directly alongside routine tasks. Dispatches multipart audio blobs to Groq's high-speed Whisper v3 API, then normalizes Devanagari/Urdu speech into clean Romanized Hinglish via Llama 3.3.
Engineering Specifications
- Whisper v3 on Groq LPU hardware (~350ms transcription)
- Few-shot transliteration engine for multilingual code-switching
- Automatic task-level attachment & optimistic note saving
Live Runtime StateVoice Latency: < 400ms
Streaming Groq Whisper v3...
“Aaj subah 6 baje gym kiya, 5km run complete ki...”