Arjun Chintapalli is co-founder and chief executive of Retriever AI, or rtrvr.ai, which builds browser agents that automate work across websites, documents and business applications. His approach turns the browser itself into an AI execution environment, using its existing authenticated sessions and structured webpages to complete tasks that conventional integrations cannot easily reach.
Chintapalli started in oil and gas, moved into software engineering at Capital One, and worked at Google from 2019 to 2024 on advertising-model infrastructure and vertical federated learning. He earned a bachelor’s degree from the University of Texas at Austin and master’s degrees in computational engineering and computer science from Georgia Tech.
He founded Retriever with Bhavani Kalisetty, its chief technology officer. The product began as a Chrome extension that accepts natural-language instructions and expanded into cloud execution, APIs, Model Context Protocol integrations and embedded agents. Chintapalli has described its evolution around several specific architectural bets:
- Browser-native authenticated context: Running inside an existing browser session lets agents use services their operators already access, without transferring passwords to a separate hosted browser. Retriever also coordinates local sessions with cloud execution.
- DOM-native browser automation: Structured webpage content gives models direct access to text and interactive elements, avoiding screenshot-heavy navigation and enabling work in background tabs.
- Parallel multi-tab workflows: Independent subtasks run simultaneously across webpages instead of accumulating failures in one extended sequence. Chintapalli demonstrated structured extraction into Google Sheets, product comparisons and browser actions across multiple tabs.
- Rover and embedded agents: Retriever’s platform extends browser automation into other companies’ products. Developers have applied it to healthcare authorization, compliance research and customized browser workflows.
Chintapalli also envisions collaborative dataset construction, with people using their own browsers to assemble information scattered across otherwise difficult-to-navigate websites, including local government and civic resources.
