2026 Scientific Symposium

Abstract

Animals display mesmerizing autonomous behaviors through complex physical interactions within their soft bodies and with the environment in which they live. These local interactions often bypass their central brain. In stark contrast, artificial robots typically rely on centralized computers that send sequential control signals to actuators. In this talk, I ask how autonomous and coordinated behaviors can be embodied directly in the physical structure of soft machines and matter. I will show soft machines with tailored structures, such as elastic shells and tubes, whose mechanical instabilities, coupled to fluid flow and pressure, give rise to nonlinear dynamical responses. These physical interactions enable complex behaviors to emerge directly from the systems’ mechanics, including memory of stimuli, synchronization of limbs, and regulated internal activity. In the final section of the talk, I will transition from machines to materials, showing how instabilities in viscous fluids can drive pattern formation in solidifying soft matter. By coordinating fluid motion via local viscous and thermal interactions, emerging global structures can be built adaptively from the bottom up.

Biography

Alberto Comoretto is a postdoctoral researcher at KU Leuven in the Department of Chemical Engineering, where he studies the emergence of patterns in viscous fluids undergoing instabilities. He received a PhD cum laude in 2025 from Eindhoven University of Technology for his doctoral research at the AMOLF institute in the Autonomous Matter Department, where he investigated physical principles for the embodiment of autonomous behaviors in soft machines. In 2024, he was a visiting researcher at the Massachusetts Institute of Technology (Department of Mechanical Engineering), where he studied how energy can be embodied in soft devices by coupling elasticity with catalytic reactions. His interdisciplinary experimental research lies at the interface of mechanics, soft matter physics, and robotics. He received an MSc degree cum laude in 2021 and a BSc degree in 2018 from the University of Trento.

Abstract

Complex biological functions emerge from the interaction of nonlinear chemical networks, transport processes, and mechanics. My research aims to translate these principles into synthetic soft-matter systems that can sense, process information, and adapt autonomously through internally coupled chemical and mechanical dynamics.
First, I will present flow-fueled reaction–diffusion systems that generate sustained, tunable spatiotemporal patterns in hydrogel devices, enabling autonomous chemical dynamics at the material scale. I will then introduce programmable DNA-based reaction networks that implement custom information-processing topologies, supported by machine learning–based analysis of complex reaction dynamics.
Extending these concepts to functional materials, I will discuss microfabricated, chemically active hydrogels, neuromorphic photonic platforms, and soft robotic systems that integrate mechano–chemical sensing, signal propagation, and actuation. I will show how chemical intelligence can be combined with elementary mechanical intelligence to achieve force-gated, mechano-adaptive material responses.
Together, these examples establish a modular sensor–processor–actuator paradigm embedded within soft matter and point toward future directions for developing fully resettable, self-regulating systems with closed-loop feedback, rewritable memory, and embodied computation.

Biography

Brigitta earned her diploma degree in Chemistry from Eötvös Loránd University in Budapest, Hungary, where she later completed her PhD in 2021 under the supervision of Prof. István Szalai. Her doctoral research focused on nonlinear reaction–diffusion pattern formation in chemical systems. Following this, she joined the Life-Like Materials and Systems Group of Prof. Andreas Walther at Johannes Gutenberg University Mainz, Germany, as an Alexander von Humboldt Postdoctoral Research Fellow. Since February 2024, she has been leading her independent international collaborative project on Next Generation Neuromorphic Soft Matter Devices, funded by the Volkswagen Foundation. Her research combines nonlinear (bio)chemical reaction networks, including synthetic DNA circuits, with chemo-responsive 3D-printed hydrogels to create autonomous information-processing metamaterials for applications in adaptive functional materials and soft robotics.

Abstract

Human skins are compliant, healable, stretchable, and multimodal soft sensors which physically govern our interactions with the world around us. By understanding the information generated by millions of distributed mechanoreceptors, we can perform tasks as diverse as threading needles, finding torches in the dark, or playing music on a violin. These capabilities are challenging to reproduce in artificial systems, requiring the close coupling of material-level sensation with higher-level contextual interpretation for intelligently-informed actions.
This talk will explore the enabling technologies behind deployable and scalable closed-loop tactile perception. This includes novel artificial skins which cover large-area complex-shaped surfaces, continuously generating thousands of streams of tactile information. Data-driven and model-based approaches interpret these multimodal signals from dynamic environmental interactions, supporting the development of technologies for force feedback during surgery, general-purpose dexterity in robotic hands, and accessible communication interfaces in prosthetic devices.

Biography

David Hardman is a Junior Research Fellow and EPSRC Doctoral Prize Fellow in the University of Cambridge’s Bio-Inspired Robotics Lab. His thesis on multimodal soft sensors reached the finals of the 2025 Georges Giralt Award for the best European PhD in Robotics, and won a CSAR Award for outstanding research with real world applications. Prior to his current position, David was a visiting researcher at EPFL’s CREATE Lab in 2024, where he worked on modular and customizable robotic fingertips.

Abstract

Today’s world is increasingly shaped by the widespread adoption of artificial intelligence (AI) across diverse domains. However, this rapid expansion is accompanied by growing environmental and energy costs. As a result, there is a demand for scalable, sustainable, and energy-efficient AI hardware capable of supporting deep neural networks (DNNs) across platforms ranging from cloud data centers to resource-constrained edge devices, without compromising performance. In-memory computing (IMC) has emerged as a promising paradigm to overcome the limitations of traditional von Neumann architectures by performing matrix-vector multiplications directly within memory arrays, thereby substantially reducing data movement and energy consumption. Nevertheless, the effective deployment of AI models on IMC accelerators requires comprehensive co-optimization across the software-hardware stack, including the entire computing hierarchy, from device non-idealities and IMC circuit design to architectural organization. This, in turn, requires holistic co-design methodologies that jointly optimize algorithmic, architectural, and circuit-level parameters to meet the diverse and evolving requirements of modern AI workloads.

Biography

Olga Krestinskaya received her Ph.D. degree from King Abdullah University of Science and Technology (KAUST), Saudi Arabia, in 2025, where she is currently a postdoctoral fellow. Her research focuses on software-hardware co-design for in-memory computing (IMC) architectures and AI hardware. She has authored several high-impact journal articles, conference papers, and book chapters, covering analog memristive neural networks, mixed-signal circuit-level implementations of IMC architectures, quantized neural networks, and brain-inspired algorithms, with a strong focus on developing energy-efficient and scalable IMC hardware for AI applications. Dr. Krestinskaya is the recipient of the 2019 IEEE CASS Predoctoral Award, the Erasmus Student Mobility Scholarship, the KAUST Dean’s Scholarship, the 2025 Web of Talents STEM Award (1st place), and multiple KAUST Dean’s Awards. Her work was recognized with the Best Poster Award at the 2nd Nature Conference on Neuromorphic Computing (2024), and she was shortlisted for the prestigious Rising Stars Women in Engineering Workshop (Asian Deans’ Forum 2024).

Abstract

Humans outperform robots in physical interaction, despite current advances in robot learning and control. AI-driven approaches are intrinsically unidirectional, constrained only by temporal causality and boundedness. Physical interaction, however, is dictated by reactive behavior governed by energy conservation, entropy, etc. Robot learning algorithms for physical interaction must therefore interface the information domain with the energy domain. In this talk, I present physics-grounded AI, a unifying framework that integrates coordinate-invariant, energy-consistent robot dynamics with imitation learning and generative AI to enable safe, stable, and human-mimetic physical interaction. Inspired by human motor control, the framework enables modular robot control via motor primitives that form a compositional control language, allowing robots to learn interaction strategies from demonstration while preserving physical consistency. I will present my recent work on diffusion-based impedance learning, which bridges generative modeling with interactive control, enabling robots to adapt stiffness and damping autonomously, even across unseen tasks and hardware configurations. The focus application of this talk is robot-assisted physical rehabilitation, where my physics-grounded AI framework allows skilled physical and occupational therapists to imprint their treatment style onto robots, enabling personalized, stable, and safe assistance.

Biography

Johannes a postdoctoral researcher in the MIT–Novo Nordisk AI Fellowship Program, working at the intersection of robotics, motor neuroscience, and AI. His research combines insights from human motor control to create robotic systems that support individuals with neurological and physical impairments. Johannes is also a Lecturer and Lab Professor for one of MIT’s flagship undergraduate Mechanical Engineering courses in Instrumentation and Measurement. Before MIT, Johannes spent nine years at KUKA, where he led research in contact-rich robotic manipulation and physical Human-Robot Interaction. As a Senior Research Engineer in Corporate R&D, he helped shape technologies that now power real-world products. His work led to ten patents, five of which were successfully transferred into industry. Johannes earned his PhD with highest distinction under the supervision of Prof. Stefano Stramigioli (University of Twente) and Prof. Neville Hogan (MIT). His thesis applied differential geometry to develop robot controllers that are safe, stable, and energy-aware, and was honored as a finalist for the European Georges Giralt PhD Award, for which only four researchers across Europe were selected. Outside the lab, Johannes mentors students in MIT Engineers Without Borders and enjoys helping them grow as future engineers.

Abstract

Building effective interactive systems has long been a central task in Human-Computer Interaction (HCI). Whether creating mixed-reality interfaces or building input devices, developers must navigate vast design spaces to identify an effective solution, which is then fixed for all users. This process relies on iterative prototyping and user testing, yet promising designs may be missed. Moreover, such “one-design-fits-all” workflows assume a single design can serve all users and contexts. This is increasingly incompatible with today’s interactive systems, where personalization and adaptation are valuable or even essential. I believe the future of HCI lies not in perfecting fixed interfaces, but in creating intelligent systems that can learn from human behavior and adapt accordingly. 
In this talk, I introduce human-in-the-loop optimization (HILO) as a unifying computational framework for this vision. In HILO, intelligent systems iteratively refine interfaces by optimizing based on observed user behavior and feedback. I will present my research contributions that advance HILO, including expanding its application scope from handling a single design objective to multiple objectives, improving its efficiency via experience learned across prior users, and developing approaches that enable scalable interaction data to further augment optimization.
Finally, I will briefly outline future directions toward establishing HILO as a foundation for future human-AI interaction. I will particularly discuss integrating HILO with generative AI to support active human-AI collaboration in creative tasks, and moving toward foundation models for design optimization that continuously learn and generalize across diverse users and interactions.

Biography

Yi-Chi Liao is a postdoctoral fellow at SIPLAB, ETH Zürich. His research lies at the intersection of Human-Computer Interaction, machine learning, and design optimization. He develops human-in-the-loop systems that adapt, optimize, and create with humans. He also serves as Associate Chair for ACM CHI and ACM UIST. Previously, he was a postdoctoral fellow at Saarland University and a research intern at Meta Reality Labs. He received his Ph.D. from Aalto University.