Speakers

Pulkit Agrawal

Massachusetts Institute of Technology, USA

Talk title: tbd

Abstract: Pulkit Agrawal’s research focuses on providing machines with human-like capabilities for manipulation and locomotion, aiming to create systems that can continuously learn from their surroundings and develop a form of common sense and physical intuition. He describes this vision as computational sensorimotor learning, a broad framework that integrates perception, control, robotics, and reinforcement learning. His work is also influenced by ideas from cognitive science and neuroscience, reflecting an interdisciplinary approach to understanding intelligence. At his lab, Agrawal investigates how robots can acquire skills through experience rather than explicit programming. A central theme is reinforcement learning, where agents learn by interacting with environments, often trained in simulation and then deployed in the real world. More recently, his research has expanded toward scalable learning systems inspired by foundation models, incorporating visual data, language instructions, and large amounts of unstructured information such as videos. Another important goal is enabling generalization, allowing robots to adapt to new tasks and environments with minimal supervision. Overall, his work contributes to the broader pursuit of embodied AI: building adaptive, versatile machines that learn autonomously and function effectively in complex real-world settings. Most recently, he has extended these principles into AI for Science, contributing to robotic platforms and machine learning methods that accelerate the discovery of novel materials and molecules.

Bio: Pulkit Agrawal is an Associate Professor in the Department of Electrical Engineering and Computer Science at MIT. He earned his Ph.D. from UC Berkeley and his bachelor's degree from IIT Kanpur and was awarded the Director's Gold Medal. His work has received multiple Best Paper Awards, the IEEE Early Career Award in Robotics, the IROS Toshio Fukuda Young Professional Award, the IIT Kanpur Young Alumnus Award, the Sony Faculty Research Award, the Salesforce Research Award, the Amazon Research Award, the Signatures Fellow Award, the Fulbright Science and Technology Award, and others.

Benjamin Alt

Talk title: Semantic Digital Twins and Virtual Robot Labs for Reproducible, Accessible Scientific Discovery

Abstract: Autonomous robots promise to accelerate scientific discovery across all disciplines of science. Using VR and semantic digital twin technologies, scientists can define experiments, program robots, and make their experiments and results available online to the scientific community. The AICOR Virtual Research building is an open-source, open-science platform for creating, hosting and accessing robotics experiments in a standardized manner, along with the data of the corresponding real-world experiments. It promises to increase the replicability and transparency of robot-based scientific inquiry, and makes robotic experimentation available to underserved communities across the globe.

Bio: Benjamin Alt is the Technical Director at the AICOR Institute for Artificial Intelligence and Co-Founder of a robotics startup building a general-purpose robot cognitive architecture. His research aims to endow robots with the abilities to competently act in the real world, while ensuring safe and human-compatible behavior. His research interests include neurosymbolic machine learning, semantic digital twins and human-robot interaction.

AICOR Institute for Artificial Intelligence, University of Bremen, Germany

Daniel Chan

Medra AI,
USA

Talk title: Physical AI, Agents, LLMs - Path Towards Autonomous Science

Abstract: Recent progress in AI for science has made it easier to generate hypotheses, design experiments, and reason over results, but a major bottleneck remains in execution. In biology, real experiments still depend on physical context that is rarely captured in software alone: instrument quirks, process observations, real-time sensory cues, tacit know-how, and failure recovery. Medra’s approach is to build Physical AI Scientists that connect scientific reasoning to modular laboratory execution, capture rich closed-loop process data, and continuously improve protocols through repeated operation in real labs. Medra frames this system around three core ingredients: modularity, measurement, and continuous learning.
In this talk, I will share Medra’s perspective on what it takes to move from narrow automation to continuous science. I will discuss why execution is not just a downstream implementation detail, but a core scientific bottleneck, and how richer logging of process data, observations, images, tactile sensing, and error handling changes what can be optimized. I will also describe why scale matters. Medra recently built a laboratory of 100 Physical AI Scientists where every run contributes to a larger learning system for experimental execution and scientific iteration.

Bio: Daniel Chan is a Member of Technical Staff at Medra, where he leads robotics on Medra Lab 1. Previously, he led deployments and applied AI. His work sits at the intersection of robotics, AI, and biology. Before Medra, Daniel worked on robotics systems at Hyundai New Horizons Studio and Amazon’s Grand Challenge Lab. He holds BS and MS degrees in Mechanical Engineering from Stanford University.

Riley Hickman

Talk title: VALIANT: Enabling Self-Learning Laboratories for Pharmaceutical Formulation Development

Abstract: Advances in machine learning and high-throughput experimentation have created new opportunities to accelerate pharmaceutical development, yet most systems remain fragmented, separating computational design from laboratory execution. Bridging this gap requires integrated platforms that connect AI-driven experiment planning with robotic workflows under real-world constraints.
In this talk, we present VALIANT, a modular, general-purpose platform developed at Intrepid Labs for pharmaceutical formulation development. VALIANT integrates AI-driven design with robotic orchestration, device control, and data management to enable closed-loop iteration across the full development lifecycle. It is designed to operate within stringent data security, reproducibility, and regulatory requirements while supporting a broad range of formulation modalities.
We describe how VALIANT is used in practice to coordinate heterogeneous laboratory systems and accelerate in vitro development timelines. We highlight results from real-world deployments, including formulations that have progressed from early experimentation through scale-up and toward clinical evaluation with industry partners.
Finally, we discuss the productionization of these capabilities into an integrated application layer that serves as a control center for formulation development, unifying experiment planning, execution, and analysis. These end-to-end systems are critical for translating advances in AI and robotics into scalable impact in pharmaceutical science.

Bio: Dr. Riley Hickman is Co-Founder and Senior Director of Engineering at Intrepid Labs, a company focused on accelerating the development of pharmaceutical formulations by integrating machine learning, robotics, and pharmaceutical science. He earned his PhD from the University of Toronto under Professor Alán Aspuru-Guzik, where his research focused on autonomous research systems for accelerating the discovery of advanced materials and functional molecules.
Prior to and through founding Intrepid, Riley has worked at the intersection of computer science, robotics, chemistry, and biology to build systems that close the loop between prediction and experiment. At Intrepid, he leads the development of VALIANT, a framework for accelerated pharmaceutical formulation development that combines machine learning, lab automation, robotics, and software into a unified platform. VALIANT is designed to operate within the stringent data security, regulatory, and reproducibility constraints of the pharmaceutical industry while enabling rapid, scalable experimentation. Through this work, Riley has helped translate AI-driven discovery systems into real-world impact, with formulations developed using VALIANT now scaled and advancing toward clinical evaluation with pharmaceutical partners.

Intrepid Labs Inc., Canada

Genki N Kanda

Institute of Science Tokyo, Japan

Talk title: The Living Laboratory: Fully Automated Science Powered by AI Agents and Physical AI

Abstract: We aim to realize a “Living Laboratory” by integrating experimental robots, AI agents, and physical AI, where biological research proceeds autonomously and scientific papers are generated continuously. While current laboratory automation mainly focuses on experimental execution, planning, preparation, and recovery (“Care”) still rely heavily on humans. We address this gap by tightly linking AI agents with physical AI to take on Care. Our approach assigns Planning Care to AI agents that translate natural language procedures into executable robotic workflows, generate consumable layouts and programs, and manage loop- and branch-rich experiments such as cell passaging. Operation Care is handled by Physical AI through a vision–language–action system that autonomously performs implicit tasks, validates programs in simulation, detects errors, and provides corrective suggestions in real time. At the Institute of Science Tokyo, we have established a centralized robotic facility integrating multiple experimental robots with VLA-enabled systems. In this talk, we will present our current progress toward the Living Laboratory, including system integration, AI-driven planning, and VLA-based operation in real laboratory environments.

Bio: Dr. Genki N. Kanda, Ph.D., PMP, is a Professor at the Institute of Science Tokyo, Medical Research Laboratory, Department of Robotic Science. He received his Ph.D. in Science from Osaka University in 2016. Since first encountering the Maholo LabDroid in 2015, Dr. Kanda has led research and implementation of laboratory automation in life sciences and regenerative medicine. In 2019, he founded the Laboratory Automation Suppliers’ Association (LASA) in Japan and continues to serve as its chairman, fostering industry–academia collaboration and community building.

Shi Xuan Leong

Talk title: tba

Abstract:

Bio:

Nanyang Technological University (NTU), Singapore

Rodrigo Moreno

REAL Lab, IT University of Copenhagen, Denmark

Talk title: Navigating Modular Reconfigurable Lab Automation: A Quick to Reconfigure Platform

Abstract: Laboratory automation equipment must be repurposed often for experiments, making the current generation of automation devices too cumbersome and complicated. However, with platforms in which modules encapsulate operations, equipment can be swapped around quickly to perform different experiments. This talk explores the state of the art of reconfigurable modular laboratory automation systems in chemistry from a roboticist perspective. A scale of modularity levels based on the hardware’s flexibility, reusability and ease of use is proposed and more specifically systems that are user reconfigurable are examined. We end by discussing the challenges and key barriers of adoption of these systems and promising research paths.

Bio: Rodrigo is an Assistant Professor working with easy to reconfigure Lab automation for chemistry synthesis processes at the IT University of Copenhagen. Rodrigo joined ITU in 2020. He holds a PhD in Systems and Computer Engineering from Universidad Nacional de Colombia. He started working in lab automation encapsulating chemical unit operations as part of the BIG-MAP EU project and nowadays continues working on reconfigurable lab automation systems with focus in nanoparticle synthesis. His research interests include AI and machine learning applied to robotics, evolutionary robotics, modular robots and automation. Other projects he is working on include the EMERGE project, an easy to build modular robot system and the MOZART EU project for developing modular surface manipulation.

Spotlight Speakers

Industry Spotlight Talk: Rich Walker from Shadow Robotics

Rich Walker

Industry Spotlight Talk: Sanket Behzad and Mohebbi Gaurav from P&G (Procter & Gamble)

Title: Toward Self-Driving Labs at P&G: Practical Lessons in Lab Automation, Robotics, and Physics-Informed AI

Abstract: Industrial R&D laboratories face need to accelerate product development while maintaining scientific rigor, reproducibility, and operational flexibility. In this talk, we will share P&G’s perspective on how robotics, AI, and scientific modeling can come together to enable self-driving laboratory workflows for faster and more reliable experimentation.
To ground this discussion, we will highlight two complementary technical directions at P&G. First, we will present a real-robot learning effort that illustrates the need for adaptive physical task execution when rigid hand-coded automation becomes brittle for complex, contact-rich behaviors. We will then connect this execution layer to a broader self-driving lab vision aimed at automated experiment design, material-property estimation, and material discovery, where physics-informed AI can play an important role. We will discuss recent work on coupling neural networks with conventional numerical solvers as one step toward more robust and practical scientific modeling for forward and inverse problems. Together, these examples illustrate how self-driving labs require both adaptive robotic experimentation in the physical world and scientifically grounded AI for guiding decisions and accelerating discovery. Lastly. we will close by outlining lessons learned from working in an industrial lab environment, as well as open research opportunities where the academic community can help advance scalable and trustworthy laboratory autonomy.

Behzad Mohebbi

Behzad Mohebbi is a Group Head at Procter & Gamble, working at the intersection of Scientific Machine Learning, Physics-informed AI, and analytical measurement science. His work focuses on developing AI-enabled digital twins by integrating advanced measurement science and machine learning for R&D applications. He received his Ph.D. in Physics from RWTH Aachen University, where his research focused on developing novel end-to-end integrated sensor-based measurement systems.

Sanket Gaurav

Sanket Gaurav is a Senior Machine Learning & Robotics Scientist at AI and Data driven Transformations, CF-R&D, Procter & Gamble, working at the intersection of robotics, AI, and autonomous systems. His work focuses on robot learning, imitation learning, reinforcement learning, and applied automation for real-world tasks. He received his Ph.D. in Computer Science from the University of Illinois Chicago, where his research focused on Teleoperation, Human Intent Prediction, and Imitation Learning methods for Collaborative Robots.

Page updated

Google Sites

Report abuse