Sowmya Rani — AI Researcher

I build AI systems that talk to humans well enough that humans don't notice. I'm a Principal Data Scientist at Verizon, working on agentic AI, multi-agent systems, LLM tooling, and the boring but critical infrastructure that keeps them safe in production. I'm also a PhD scholar in Computational Linguistics at BITS Pilani (Goa), which is really just a fancy way of saying I spend a lot of time thinking about why language models still get language wrong. Lately, I've been stuck on one problem: customer support that doesn't make people want to throw their phone across the room. Most support chatbots are one tired agent doing ten jobs. I think the real answer looks more like a team: triage, specialist, escalation, memory, each one good at its own thing, built on frameworks like Google ADK. I write about what works, what breaks, and what nobody warns you about in the docs. Here's what you'll find on this profile. Posts on multi-agent systems, ADK, and agent evaluation. Lessons from shipping AI inside a Fortune 50 telco, which is very much not the same as shipping a demo. And every so often, some honest writing about how to keep growing as an engineer when the field reinvents itself every six months. If you're working on customer-facing AI, agent evaluation, or you just want to argue about whether agents are real engineering or fancy plumbing, my DMs are open.

Stackforce AI infers this person is a Data Science and AI expert specializing in multi-agent systems and customer support technologies.

Location: Bengaluru, Karnataka, India

Experience: 16 yrs 1 mo