What does Ansh Singhal build?
Ansh Singhal builds production AI and backend systems focused on retrieval-augmented generation, realtime inference, and distributed reliability. This project index summarizes three core case studies with measurable outcomes: DHARA retrieves answers from over 1 million indexed documents while maintaining sub-100ms p95 retrieval latency; VoiceOps AI runs realtime speech ingestion and transcription at 50ms latency with 1000+ concurrent sessions; and Flood-GAN synthesizes high-resolution flood imagery at 1024px with a reported FID of 28.4. The projects are presented as implementation-first engineering narratives covering problem framing, architecture choices, and operational outcomes. Together they represent a practical portfolio of systems engineering work where latency, throughput, and answer quality are treated as first-class production constraints rather than demo-only metrics.