Bhavani Kalisetty

Bhavani Kalisetty is co-founder and chief technology officer of rtrvr.ai, building browser agents that research websites, navigate software, and turn scattered online information into completed work. Her approach gives agents direct access to webpage structure and existing user context, allowing them to operate across authenticated sessions, documents, and multiple tabs.

Kalisetty studied computer science at the National Institute of Technology, Warangal, and earned an MBA from the University of Southern California’s Marshall School of Business. At Adobe in India, she developed artificial-intelligence systems for abuse detection. She later managed post-sales work for Google Ad Manager publisher accounts and led a publisher-focused generative AI task force within Google Ads before founding rtrvr.ai with Arjun Chintapalli.

Their initial product, Retriever, transformed a Chrome side panel into a natural-language interface for browser-based work. At AI Engineer Summit 2025, Kalisetty demonstrated document summarization, competitive research, financial analysis, custom integrations, and chart generation, including an agent that recovered from an incorrect starting URL to find the requested financial data.

• DOM-native browser agents: Kalisetty favors reading a webpage’s underlying structure over repeatedly interpreting screenshots, reducing ambiguity while enabling background-tab processing and direct interaction with page elements.

• Automation inside existing user context: Agents can use documents, signed-in sessions, and personalized information already available in the browser without requiring users to share passwords with a separate cloud-hosted browser.

• Multi-tab structured extraction: Retriever can navigate across websites, derive a schema from a natural-language request, collect specific information, calculate additional fields, and deliver usable results in Google Sheets.

• Rover and website-native agents: Rover embeds an agent directly inside software products to guide users through workflows and complete actions. Kalisetty argues that websites should actively participate in agentic interactions, giving embedded agents access to live page structure.

• Task completion as the meaningful benchmark: Her comparison of browser-agent products emphasizes completed assignments, factual accuracy, execution visibility, and the cost of successful results—priorities also reflected in rtrvr.ai’s browser-agent data and evaluations.