Speaker
Description
The agentic AI ecosystem is maturing, but authentication and authorization remain its weakest link. MCP evolved from no auth model to mandatory OAuth 2.1 with PKCE and step-up authorization within an year.
Agent frameworks like Google ADK, LangGraph, and LangChain handle identity and consent differently. Yet critical gaps persist - machine-to-machine flows are not addressed, token propagation through multi-agent chains has no standard, and the fundamental question remains: When no human was in the loop, who authorized what?
This talk examines what MCP and agent frameworks get right, where they fall short, and the hard problems teams face when deploying authenticated agents in production.
Ideal for backend developers, and anyone building agentic AI systems.
Session author's bio
I’m Swaraj Pande (スァラジョ・パンデー) - a Software Engineer (L2) at Red Hat, a proud RHCA (Red Hat Certified Architect), and an open-source advocate. I specialize in building scalable cloud-native architectures, automating complex infrastructure, and driving AI/ML innovations.
At the heart of my entire tech stack is Python. It is the essential bridge I use to connect infrastructure, massive data workflows, and artificial intelligence into seamless, production-ready systems. I'm also deeply passionate about LLM orchestration, RAG pipelines, AI observability, and designing context-aware agents using frameworks like MCP.
I'm also a passionate tech advocate and frequent speaker at Delhi-NCR and other tech meetups happening in India, where I share insights on Python, cloud-native technologies and the evolving AI landscape.
Always open to connecting with fellow engineers, open-source enthusiasts and innovators!
| Social Media | https://x.com/swarajpande05 |
|---|---|
| In Person Attendance | Remote |
| Agree to Privacy Policy and Notice | I agree |
| Please confirm that there are included headshots of all speakers in their profiles | Yes |
| Level of Difficulty | Intermediate |