We are delighted to welcome you to our AIchemy Hub’s monthly webinar series.
This month’s talks:
Dr. Sterling Baird – So, you want to build a self-driving lab?
The scientific community has been exploring how to utilize AI and automation to create “self-driving” (i.e., autonomous) scientific laboratories with the goal of accelerating the rate of scientific discovery (i.e., new materials and new phenomena). Like peering over the edge of a high-dive, getting into the space of self-driving labs, or even AI and automation in general, can be daunting. Recently, self-driving labs for chemistry and materials science have led to accelerated scientific discoveries related to climate change, energy, and medicine.
Each of these labs are usually built over the course of several years by interdisciplinary teams, requiring capital in the millions. Of the reported self-driving labs, many have only ever initiated a single “glory flight” campaign. Many naturally find themselves asking questions such as the following. What exactly is a self-driving lab? Where do I start? What are the benefits and risks? While the answers are highly dependent on circumstance and project visions, we provide perspectives on the hardware, software, personnel, and other requirements necessary to be sustainably successful in this space. This will cover topics such as robotics, computer vision, machine learning, workflow orchestration, and “frugal twins”.
He will also describe the training efforts within the Acceleration Consortium, intended to reduce the barrier-to-entry for and de-risk the adoption of self-driving labs through a range of online microcourses and in-person training experiences.
Dr. Austin Mroz – Multi-Fidelity Bayesian Optimization in Chemistry: Open Challenges and Major Consideration
Multi fidelity Bayesian optimization (MFBO) leverages experimental and or computational data of varying quality and resource cost to optimize towards desired maxima cost effectively. This approach is particularly attractive for chemical discovery due to its ability to integrate multiple information streams. Here, we investigate the application of MFBO to accelerate the identification of promising molecules and materials. We specifically explore the conditions under which lower fidelity data can enhance performance compared to single-fidelity problem formulations. We then discuss the utility and accessibility of this powerful optimisation framework and introduce a web application to guide users through implementing MFBO in their workflows.
Following the presentations, there will be time for questions from the audience.
Sterling Baird
Director, Training and Programs
Austin Mroz
Postdoctoral Fellow
