Title:Mapping Biology with AI to Revolutionize Drug Discovery Abstract:The persistent challenge of "Eroom's Law", the decades-long decline in drug discovery R&D efficiency, necessitates new paradigms for biological research. This talk presents a high-throughput, data-driven approach centered on generating and modeling petabyte-scale morphological datasets. Using an automated platform, millions of experiments are performed weekly, applying diverse perturbations (e.g., small molecules, whole-genome CRISPR knockouts) to human cells, with phenotypic responses captured via high-content microscopy using the Cell Painting protocol. A key focus of this talk is the development and scaling of representation learning models to interpret this massive image data. We will detail the technical progression from weakly-supervised learning (WSL) to large-scale, self-supervised foundation models, specifically Masked Auto-encoders (MAEs). We will demonstrate that increasing model size and data volume yields more powerful representations that effectively recapitulate known biological relationships, creating robust "maps of biology." The practical utility of these models will be explored through case studies, including the identification of novel biological insights and the ability to uncover subtle, large-scale data artifacts, such as "proximity bias" in CRISPR-based functional genomics screens. Finally, we will discuss future directions focused on building a more holistic, multi-modal understanding of cellular states by integrating orthogonal datasets with these rich morphological profiles. Speaker:Safiye CelikDirector of Data Science, Recursion Bio:Safiye Celik is the Director of Data Science at Recursion, with 18 years of experience in computing and AI, including 12 years leading high-impact initiatives in industry and government research. She holds a PhD in Computer Science and Engineering from the University of Washington, where her research applied machine learning to characterize complex biological systems, uncovering novel insights in cancer and Alzheimer's disease. At Recursion, she leads a team that develops, deploys, and optimizes machine learning models to transform experimental data into comprehensive maps of biology, advancing the vision of industrializing drug discovery. Host: Professor Ulkuhan Guler

These trainings are divided into a 101- and 201-level and will be offered on a termly basis in Stratton Hall 311. Both trainings will include creating an action plan and discovering new resources for continued learning. Please contact Lauren Feldman (they/she) at diversity@wpi.edu with any questions or accommodations requests. All are welcome to attend! 101: Learn about identities, pronouns, and how to advocate for your students and colleagues. A-Term: Wednesday, September 10, 2:00-3:30pm ET Zoom: Tuesday, October 14, 1:30-3:00pm ET B-Term: Tuesday, October 28, 11:00am-12:30pm ET 201: Learn about LGBTQIAP+ history, queer identity in other cultures, and queer liberation. Participants are encouraged to either attend the 101 training or have solid foundational knowledge before attending the 201 training. A-Term: Wednesday, September 17, 2:00-3:30pm ET B-Term: Wednesday, November 5, 1:30-3:00pm ET

On Wednesday, November 5, from 3:30–4:30 PM in the i3 Lab, join us for a panel discussion featuring rotational and leadership development program coordinators who manage top programs across many industries. Learn how these programs accelerate professional growth, build leadership skills, and open doors to diverse career pathways in today’s competitive business world.

Live Stats

Go Behind the Scenes at Polar Beverages!On Thursday, November 6, join us for an exclusive site visit to Polar Beverages, Worcester’s iconic company with a rich history of branding, innovation, and community impact.The bus departs at 10 AM and returns by 12 PM, giving participants a unique opportunity to see firsthand how Polar has built its reputation as a leader in the beverage industry. Learn about the company’s operations, explore its production process, and gain insights into how tradition and innovation come together to shape a lasting brand.This site visit is a highlight of Business Week, don’t miss the chance to experience a Worcester business legacy up close!Registration for this event is limited and preference will be given to students attending as part of a course assignment or experience. Click this link to register today!

Semidefinite Relaxations for Robot Perception and Control: From Theory to Practice and Back Abstract:Many problems in robot perception, control, and planning can be formulated as nonconvex polynomial optimization problems (POPs). The Moment–SOS (sums-of-squares) hierarchy provides a principled approach by relaxing a nonconvex POP into a sequence of convex semidefinite programs (SDPs) whose optimal values converge to that of the original problem. Despite its theoretical elegance, the framework is often viewed as impractical due to the need to solve large-scale and ill-conditioned SDPs.In the first part of this talk, I will show how problem-specific structure in robotics can be exploited to make Moment–SOS relaxations practical at scale. On the perception side, I will present XM, our structure-from-motion pipeline that leverages monocular depth prediction to cast bundle adjustment as a polynomial optimization problem. Using GPU-accelerated low-rank Riemannian optimization, XM solves bundle adjustment instances with thousands to tens of thousands of images. On the control side, I will introduce SPOT, a trajectory planning pipeline that exploits sparsity inherent in robot motion planning problems, such as Markov and kinematic chain structure. With a GPU-accelerated ADMM solver, SPOT is able to generate near–globally optimal trajectories within seconds, including contact-rich motions.In the second part, I will turn to theory, briefly highlighting our recent work on understanding and accelerating the convergence of first-order methods for SDPs. This includes proving local linear convergence of ADMM, designing a composite polynomial filter for approximate projection onto the positive semidefinite cone, and analyzing the slow-convergence regions of ADMM.Bio:Heng Yang is an Assistant Professor of Electrical Engineering in the School of Engineering and Applied Sciences (SEAS) at Harvard University. He received his Ph.D. from MIT in 2022 and his B.S. from Tsinghua University in 2015. He leads the Harvard Computational Robotics Group, which is broadly interested in the intersection of theory and practice, with a focus on computational algorithms that are robust, efficient, and equipped with strong performance guarantees. His work has been recognized with multiple awards, including the Best Systems Paper Award at RSS 2025, a 2025 Best Paper Award Finalist from the IEEE Technical Committee on Model-based Optimization for Robotics, a Best Paper Award Finalist at RSS 2021, the Best Paper Award in Robot Vision at ICRA 2020, a Best Paper Award Honorable Mention from IEEE Robotics and Automation Letters in 2020, and recognition as an RSS 2021 Pioneer.