Super Data Science: ML & AI Podcast with Jon Krohn 941: Multi-Agent Human Societies, with Dr. Vijoy Pandey
59 snips
Nov 18, 2025 
Vijoy Pandey, an engineering researcher and head of OutShift at Cisco, envisions a society where humans and AI agents collaborate on scientific discoveries and physical tasks. He discusses his work on AGNTCY, an open-source platform designed for agent-human interaction. Vijoy delves into the importance of trustworthy AI, the challenges of traditional identity management, and the future of access control with T3BAC. He emphasizes the need for open standards and community involvement to ensure smooth integration into various systems.
AI Snips
Chapters
Books
Transcript
Episode notes
Drug Discovery Pipeline Example
- A pharma example shows a pipeline: LLMs for planning, foundation models for exploration, and robotic wet labs for real-world testing.
- Additional agents handle efficacy, cost, compliance and link all stages into one workflow.
Agency Provides The Agent Infrastructure
- Agency is an open-source framework to scaffold agent discovery, identity, composition, communication, and evaluation across vendors.
- It embraces multiple protocols (like MCP, A2A) and aims to prevent fragmentation early.
Agents Need A TCP/IP-Like Scaffold
- Today's agent ecosystem is like pre-TCP/IP internet: islands of incompatible systems need standardization and scaffolding.
- Discovery (like DNS), identity, messaging, and evaluation form the essential scaffold for agents.
