Dr. Christopher D. Stiles is Chief Scientist of the Electrical and Mechanical Engineering Group at the Johns Hopkins Applied Physics Laboratory (APL), where he shapes strategy and leads advanced research across materials, computation, and autonomy. He earned dual B.S. degrees in Physics and Mathematics and a Ph.D. in Nanoscale Science and Engineering from the University at Albany.
Dr. Stiles develops computational methods and AI tools that accelerate scientific discovery, including AI-enabled exploration of novel superconductors and closed-loop, data-driven pipelines linking prediction, synthesis, and measurement. His work spans multiscale modeling, AI, high-performance computing, computational physics, quantum engineering, advanced manufacturing, chemistry, biology, and in situ resource utilization. At APL, he builds cross-disciplinary teams to translate foundational advances into mission capability. He is also an Assistant Research Professor in Mechanical Engineering at Johns Hopkins University and Vice-Chair of the Mechanical Engineering program in Johns Hopkins Engineering for Professionals, where he mentors emerging leaders.
Presentation Title:
Choosing Wisely: Paths Toward Insight-Maximizing Autonomy in Materials Discovery
Presentation Abstract:
Autonomous laboratories promise accelerated materials discovery, but practical closed-loop operation reveals a central lesson: the hardest challenge is often not choosing the next experiment, but coordinating the evolving web of predictions, simulations, literature, characterization data, manufacturing constraints, and expert decisions that define each action. This talk presents lessons learned from constrained AI-driven closed-loop materials discovery and manufacturing workflows spanning lunar ISRU materials to terrestrial wear-resistant materials. As generative models, physics-based tools, high-throughput characterization, and additive manufacturing entered the loop, the system increasingly depended on unstructured and semi-structured knowledge that conventional workflow tools could not easily capture. We describe how a deep-agent research environment was used to interact with this knowledge, connect analyses, preserve task context, and support reusable handoffs across people, agents, and computational workflows. These lessons move autonomous laboratories toward systems that are faster, more traceable, adaptable, manufacturability-aware, and insight-generating with each action.