Between Looking Right and Actually Working: Lessons From Building AI Agents for Robotics and Chip Design
Session Overview
AI agents can deliver remarkable results in controlled demonstrations, but the moment they are deployed in real workplaces or physical environments, they encounter a very different class of problems. Robots must behave reliably in unpredictable real-world settings, chip-design agents must meet demanding standards for accuracy and verifiability, and enterprises must operate AI continuously despite incomplete data and complex legacy systems.
In this panel, three experts who have led work in robot learning, AI-powered chip design, and enterprise data and AI operations will examine the difference between AI that appears to work and AI that can actually be trusted and used in production.
Moving beyond model performance alone, the discussion will explore the practical conditions required to deploy agents in the real world, including data quality, evaluation frameworks, human oversight, system integration, failure recovery, and accountability.
The panel will pay particular attention to high-stakes domains such as robotics and semiconductor design, where the cost of failure is significant. It will examine how much autonomy agents should be given, whether multi-agent systems genuinely improve complex problem-solving, and what organizations must prepare before moving from pilots to production.
Key Discussion Points
- Reliability beyond the demo: why strong benchmark or prototype performance does not guarantee successful real-world deployment
- Evaluation and verification: how robotics, chip design, and enterprise environments define acceptable performance and failure
- Human oversight: where people must remain involved in high-risk and high-cost decisions
- Multi-agent systems: when collaboration among agents adds value and when it creates additional complexity
- From pilot to production: the data, infrastructure, governance, and operational capabilities organizations need first
Speakers
Dr. Pannag Sanketi is the founder of Avolla, an early-stage robotics and AI startup, and most recently served as Robotics Lead at Google DeepMind.
During his time at DeepMind, he co-founded and led Open X-Embodiment (RT-X), one of the largest collaborative robot-learning initiatives to date, and contributed to the development of RT-2.
He has also built a range of advanced robotic systems, including the world’s first agile robot to reach amateur-level human performance in table tennis. His work spans large-scale robot learning, embodied intelligence, and real-world autonomous systems at the frontier of Physical AI.
He earned his Ph.D. from UC Berkeley and his bachelor’s degree from IIT Madras.
Michael Albada is a Platform Architect at NVIDIA, where he focuses on accelerating hardware and chip design through agentic AI.
He specializes in the design, evaluation, and production deployment of AI agents and multi-agent systems, with expertise spanning large language models, reinforcement learning, orchestration, and AI evaluation.
Previously, he served as a Principal Applied Scientist at Microsoft, contributing to planning, orchestration, and skill-selection systems for Microsoft Security Copilot. He has also built large-scale machine-learning systems at Uber, ServiceNow, and several startups.
He is the author of Building Applications with AI Agents, published by O’Reilly Media. He holds a master’s degree in Computer Science from Georgia Tech and an MPhil in Public Policy from the University of Cambridge.
Prerit Mishra is a data and AI leader at DHL. As Director of Data Quality, he is responsible for building trusted data foundations that enable AI to deliver reliable impact across large-scale global operations.
He previously served as DHL’s Head of Data and AI for Asia Pacific, where he built and scaled regional data and analytics capabilities and helped embed AI into day-to-day business operations.
Across more than 15 years at Micron and DHL, he has worked at the intersection of data, analytics, and AI in manufacturing and supply-chain environments, with a focus on moving enterprise AI from pilots into production.
He is an active speaker across Asia Pacific on enterprise AI adoption, human-AI collaboration, and the operational impact of AI. He holds a master’s degree in Automation from the National University of Singapore.