Prakhar Goel builds resilient real-time infrastructure at Adalat AI, a legal-tech nonprofit working to reduce court delays in India. He focuses on making systems survive unreliable networks in high-stakes environments where software errors have real consequences.
His work tackles the messy reality of live-transcription AI. He frequently shares practical lessons on handling network failures, breaking down how to design server-side protocols that recover websocket streams perfectly using replay buffers, backpressure, and sequence numbers.