divya Avanigadda

I build AI systems that work at scale. Over the past years, I've designed production systems handling thousands of concurrent requests, orchestrated multi-agent LLM workflows, and shipped Gemini-powered document intelligence to production. I'm particularly interested in the hard problems: how do you make AI systems reliable, explainable, and trustworthy? How do you orchestrate agents without token explosion? How do you know when a model shouldn't predict?