Amol Kapoor is chief executive of Nori Agentic, which builds AI agents that work with an organization’s code, documents, and operational knowledge. A former Google Research scientist and co-founder of SOOT, he designs systems around a practical conviction: AI works better when its tools and representations match how models process information.
Kapoor studied at Columbia University and worked on graph neural networks, computer vision, three-dimensional reconstruction, and robotics at Google Research. He co-authored research on multi-scale graph convolution and contributed to work reconstructing city models from historical imagery. He later co-founded SOOT with Jake Harper and served as chief technology officer, developing an AI-powered visual filesystem that organized creative material through visual similarity, color, objects, and concepts.
At Nori, Kapoor applies that understanding of structured information to agents operating within companies. Nori Sessions connects organizational knowledge to workflows including internal questions, documentation, operational responses, and presentation development.
- HTML as an agent-native visual medium. Kapoor argues that models struggle with graphics when forced into human-oriented canvases or coordinate-heavy SVG. HTML and CSS offer semantic structures—headings, grids, typography, and layouts—that agents can manipulate while browsers handle rendering. His AI Engineer presentation demonstrates the approach across branded slides, documents, video, and PDFs.
- Company context before visual polish. Useful board decks and sales materials need substantive organizational information alongside strong design. Kapoor connects agents to company materials so they can generate both content and presentation, then incorporate spoken feedback into an HTML-to-PDF workflow.
- SPACE Development. His framework—Search, Plan, Assert, Code, Evaluate—organizes agent work around gathering context, defining success, executing, and checking results. Kapoor applies this structured reasoning process beyond programming to broader knowledge work.
- Portable skills with human oversight. Nori Skillsets packages reusable agent behaviors across coding platforms. Kapoor favors reviewable configurations, targeted agents, task-specific evaluation, and human judgment over unmanaged skill proliferation or fully unattended software factories.
