Dr. Sreekanth Pannala
Director, Strategic Partnerships, Public Sector and Sovereign AI AMDAbstract
Glass-Box AI for Industry: Physics-Conditioned Digital Twins and Agentic Discovery on AMD
AI agentic systems are starting to plan studies, call simulators, and decide what to try next. For regulated industries such as petrochemicals, oil and gas, mining, and manufacturing need a higher standard of evidence. Before a recommendation gets implemented on a reactor or a reservoir, engineers need to know what the model assumed, what it measured, and where it stops being reliable.
This talk presents a glass-box approach to industrial digital twins rather than a black-box. Each twin starts from a reduced physics model with physically meaningful parameters, is calibrated against high-fidelity simulation on HPC (or data from experiments where available), and adds a small neural network for the residual only where it improves on the physics out of sample. The twin reports how much of each answer comes from physics, how much is learned, and how uncertain it is.
Agents built on these twins turn a plain-language engineering goal into a checked workflow with a record of every step. The talk shares measured results, including a hypothesis that testing refuted, and shows five live research prototypes: an agentic discovery loop, an oil and gas lifecycle twin, a sustainable aviation fuel process chain, a fusion plasma twin, and an additive manufacturing twin.
It closes with an invitation to Saudi industry and universities to test these methods on their own workloads, on AMD platforms from workstation to cluster.
Biography
Sreekanth Pannala is Director of Strategic Partnerships, Public Sector and Sovereign AI at AMD, working with governments, national laboratories, universities, and industry on AI and HPC for science and engineering.
Before AMD he was a Research Fellow at SABIC, where he invented the ANJEVOC reactor and initiated the BLUEHERO platform, and he spent 16 years at Oak Ridge National Laboratory, rising to Distinguished Research Staff. His work spans multiphase flow, reacting systems, batteries, and physics-conditioned AI for digital twins in energy and manufacturing.
LinkedIn: linkedin.com/in/pannala
Attend this session
The Saudi HPC/AI Conference runs 26–28 October 2026 at KAUST. Attendance is free, but registration is required.
