Vishnu R Potti

I’m a Product Engineer (AI) focused on building production-grade systems for customer support and sales intelligence. I design and deploy LLM-powered retrieval and agent-based solutions that serve 5,000+ daily users and handle 10,000+ queries, with a strong emphasis on multi-agent orchestration and RAG pipeline design.

At Commvault Systems, I re-architected an underperforming retrieval system by benchmarking and redesigning semantic reranking, context-aware chunking, and vector search optimization—improving accuracy from ~85% to ~95%. I also reduced average response latency by ~30% using retrieval, prompt optimization, and parallelization, and integrated support bots and agents into Slack for internal workflows.

Beyond retrieval, I’ve built end-to-end competitive intelligence and autonomous troubleshooting agents, including an MCP-based diagnostic MVP that reduced median diagnosis time from days to 20–30 minutes. I further drive reliability and safety with red-teaming automation aligned to OWASP LLM Top 10 and manage cost through prompt engineering, token management, and response caching; through freelancing, I’ve also shipped AI call intelligence and startup discovery/enrichment multi-agent systems.