At Amazon Web Services, I partner with startups and engineering teams to take AI systems from technical discovery through production. I diagnosed GPU serving bottlenecks for an AI-native startup, cutting model cold-start latency by 87% and helping reverse a multi-cloud evaluation.
I also helped productionize an agentic voice AI platform for India’s largest digital freight marketplace, combining multilingual streaming ASR/TTS, Amazon Bedrock, and EKS. It achieved under 1.5-second p90 voice latency and reduced operating costs by 78%.
Previously, at Unifo Private Limited, I architected and built an intelligent document processing platform using RAG and LLM reasoning, replacing ABBYY and Nanonets. I also built a RAG-based customer-support assistant that autonomously resolved 50% of inbound requests.
