At QuantumBot Private Limited, I build production-grade, multi-tenant AI systems across WhatsApp and email. I cut AI inference costs by 80% and increased prompt-cache hit rate to about 89% through tenant-aware routing, prompt caching, and token metering.
I also built multimodal WhatsApp intake for voice notes, images, and PDFs, and shipped visual catalogue search using Gemini image embeddings and Qdrant. For pharma sales forecasting, I benchmarked three model families and achieved 24% SMAPE with Prophet.
