Inside the Agentic AI Summit 2025: UC Berkeley’s Historic Gathering

Still buzzing in my ears from the incredible The Agentic AI Summit 2025 took place at UC Berkeley on August 2nd, 2025. Attended Agentic AI Summit at UC Berkeley last Saturday, 2nd August 2025. This was a significant event hosted by the Berkeley Center for Responsible Decentralized Intelligence (Berkeley RDI). 2,000+ joined in person, 30,000+ tuned in online. The energy was electric, visionaries, engineers & researchers shaping the future of Agentic AI. The event is organized into 3 groups 1. Main Stage 2. Workshops 3. Frontier Stage

Key details about the summit: Event Overview: The summit built on the momentum of Berkeley’s popular LLM Agents MOOC series. It was billed as the first summit dedicated to Agentic AI in the Bay Area and aimed to be the largest gathering of its kind with 1,500+ attendees. 2025 is a Year of Agents.

Technical Focus: The event brought together leading researchers, engineers, and industry leaders to discuss the future of intelligent, autonomous agents, with an emphasis on real-world deployment over hype agentic ai summit 2025. Research topics in Frontier Stage included Agent Architecture, AI Safety & Alignment, Reasoning & Planning, Benchmarks & Evaluation, Fundamental Capabilities, and Multi-agent & Decentralization AgentX. I found these talks especially fascinating: on multi-agent orchestration and key insights on multi-turn reinforcement learning on reasoning, thinking fast and slow on RAGEN’s approach to agent RL training on safety alignment and guardrails There were a ton of other awesome topics put together for such a thought-provoking event. New benchmark paper MedPAIR by @YuexingHao won the second place in the AgentX competition and first place in the Schmidt Science multi-agent competition. Yotta Labs is working closely with researchers from UC Berkeley on cost efficient Agent development and Multi-Agent orchestration.

My other stories

System design interviews are a great way to assess a candidate’s problem-solving skills. This looks like a helpful resource for anyone preparing for roles involving ***ML system design and ML inference interviews.***

The summit sessions covering technical aspects of AI agent development, with particular attention to optimization, multimodality, developer experience, and security integration. Also I enjoyed poster sessions and amazing stalls from PyTorch, Writer, Cisco, AMD, NOBELERA and many more. Great sharing from Vinod Khosla at UC Berkeley Agentic AI Summit

Overflowing rooms in PyTorch Foundation workshop on “Architecting an Open Agentic Future” packed 125+ people into a room for 75, spilling into hallways & rooftops.

🟢 Watch Summit Recordings ***here ***Had a wonderful time at @BerkeleyRDI Agentic AI Summit! Excited to see where Agentic AI goes next!