Tianqi Chen is an Assistant Professor at the Machine Learning Department and Computer Science Department of Carnegie Mellon University (CMU), with a courtesy appointment as a Professor in the Electrical and Computer Engineering Department within the College of Engineering. His research focuses on scalable machine learning systems, compiler optimization, and efficient deep learning frameworks. He holds a PhD from the Paul G. Allen School of Computer Science & Engineering at the University of Washington. Key contributions include the creation of XGBoost, Apache TVM, and MLC-LLM—widely adopted systems for machine learning and large language models. His work bridges algorithmic innovation with high-performance computing, emphasizing efficient deployment, quantization, and edge computing. Recent publications highlight advancements in LLM serving (e.g., WebLLM, Flashinfer), compiler-driven optimizations (e.g., TVM, Relax), and low-latency inference techniques (e.g., Magicdec, Tilus). These efforts address scalability, energy efficiency, and cross-platform compatibility in modern AI systems. Chen’s research has been applied to diverse domains, including music AI, browser-based inference, and microservice architectures for LLMs. His work underscores the importance of system-level thinking in advancing AI capabilities.
Yee Whye Teh is a Professor at the Department of Statistics, University of Oxford, and a research scientist at DeepMind. His work focuses on statistical machine learning, including probabilistic learning, Bayesian nonparametrics, deep learning, and Monte Carlo methods. He co-directs the ELLIS programme on Robust Machine Learning and has held roles such as Programme Co-chair for ICML 2017. Teh has delivered keynotes at UAI 2019, an IMS Medallion Lecture at JSM 2019, and the Breiman Lecture in 2017. His research emphasizes scalable inference algorithms, hierarchical models, and applications in genetics and natural language processing. Teh's educational background includes a PhD from the University of Toronto (2003) and a Master's from the same institution (2000). He has contributed to widely used software tools like the Sequence Memoizer and has been recognized for his work through prestigious lectureships. Research interests span Bayesian nonparametric models, MCMC methods, and their applications in genetics and data compression. His lab collaborates on projects like fragmentation-coagulation processes for genetic variation modeling and Mondrian forests for online learning. Teh advises students through Oxford's graduate programs, though he notes high demand for mentorship. His work often bridges theory and practice, addressing challenges in big data learning and small data problems.
Benjamin Van Roy is a Professor at Stanford University since 1998, affiliated with the Departments of Electrical Engineering and Management Science and Engineering, and the Institute for Computational and Mathematical Engineering. He leads the Efficient Agent Team at Google DeepMind and previously held leadership roles at Morgan Stanley, Unica, and Enuvis. He holds SB, SM, and PhD degrees in Computer Science and Electrical Engineering from MIT, advised by John Tsitsiklis. His research focuses on reinforcement learning, alignment, and information theory, with contributions to machine learning foundations, optimization, and finance. He has authored over 150 publications, including influential works on Thompson Sampling, approximate dynamic programming, and exploration strategies. His honors include INFORMS and IEEE Fellowships and the INFORMS Lanchester Prize. Van Roy advises doctoral students across academia and industry, with graduates at top institutions and companies like Meta, Tesla, and Citadel. He teaches courses on reinforcement learning, stochastic control, and optimization. His open-source projects include Epistemic Neural Networks and the Neural Testbed for evaluating machine learning models. Key contributions include foundational work in reinforcement learning theory, scalable methods for recommendation systems, and applications in finance and resource allocation. His research bridges theoretical insights with practical applications, emphasizing alignment and safety of AI systems.
Renjie Feng is a Research Fellow in Mathematics and AI at the School of Mathematics and Statistics and the Sydney Mathematical Research Institute , University of Sydney. His work bridges probability theory, statistics, and applications in machine learning, deep learning, and artificial intelligence. His research interests focus on probability theory and its applications to machine learning , random matrix theory , and statistical physics . He investigates extreme value problems, spectral properties of random matrices, and topological features of random fields over Riemannian manifolds. Recent publications highlight trends in random matrix theory (GUE, GOE, GSE), extreme gap problems , determinantal point processes , and Wiener chaos . Collaborative works with F. Götze, D. Yao, and R. Adler emphasize U-statistics , multivariate linear statistics , and random topology inspired by Poisson point process studies.
Jackie Chit Kit Cheung is an Associate Professor in the School of Computer Science at McGill University, where he co-directs the Reasoning and Learning Lab. He holds the Canada CIFAR AI Chair and serves as an Associate Scientific Co-Director at the Mila Quebec AI Institute. He is also a consulting researcher at Microsoft Research Montreal. Academic Background: Ph.D. in Computer Science, University of Toronto (2010–2014) M.Sc. in Computer Science, University of Toronto (2008–2010) B.Sc. (Honours) in Computer Science, minors in Linguistics and German, University of British Columbia (2004–2008) Research Interests: Jackie Cheung's research lies at the intersection of natural language processing, machine learning, and cognitive science. He focuses on natural language generation , automatic summarization , commonsense and pragmatic reasoning , and the evaluation of NLP systems . His work aims to build language models that reflect real-world structure and reasoning, with applications in health, education, and language revitalization. He is particularly interested in how implicit meaning is processed in context, a core concern in linguistic pragmatics. Publication Trends: His recent publications (2024–2025) show a strong emphasis on evaluation methodologies (e.g., COSMIC, ECBD), hallucination and factuality in language models, coreference and reasoning , and long-context modeling . He frequently collaborates across institutions and integrates insights from linguistics and psychology into NLP system design and analysis. Scientific Awards: Best Paper Award, ACL 2018 Outstanding Paper Award, NAACL 2018 Student Research Workshop SAC Award, ACL 2024 (for COSMIC) Best Poster Award, CMDO AI-Health Symposium 2024 Advising and Grants: He advises a large and diverse group of graduate students, including PhD and Master’s candidates, often in co-supervision with other faculty. His group has received support from major AI and health research initiatives, including CIFAR and Mila. He has trained alumni who have gone on to faculty and industry research positions. He has served in leadership roles in top NLP conferences, including as Senior Area Chair, Workshop Chair, and Program Chair of Canadian AI 2018. Labs and Teams: He co-directs the Reasoning and Learning Lab at McGill and is deeply involved with Mila – Quebec AI Institute . He founded the NLP Reading Group at McGill, which brings together researchers from computer science, linguistics, and information studies to discuss theoretical and applied NLP topics.
Michael A Osborne is Professor of Machine Learning at the University of Oxford and leads the Bayesian Exploration Lab . He serves as Director of the EPSRC Centre for Doctoral Training in Autonomous Intelligent Machines and Systems and co-directs the Oxford Martin AI Governance Initiative. His research focuses on Bayesian optimization, Gaussian processes, and probabilistic numerics with applications in quantum devices, battery modeling, and AI governance. Key Positions: Professor of Machine Learning, University of Oxford Official Fellow, Exeter College Co-founder of Mind Foundry Lead Researcher, Oxford Martin Programme on Technology and Employment Research Themes: Probabilistic modeling for quantum systems Uncertainty quantification in energy storage AI safety and societal impact analysis Automated experimental design Quantum device calibration Probabilistic numerical methods Technical Contributions: Bridging reality gap in quantum devices Efficient Bayesian quadrature techniques Personalized neurostimulation algorithms Automated measurement protocols Quantum-classical hybrid ML
Celestine Mendler-Dünner is a Principal Investigator at the ELLIS Institute in Tübingen, co-affiliated with the Max Planck Institute for Intelligent Systems and the Tübingen AI Center. She leads the Algorithms and Society research group, focusing on machine learning in social contexts and the role of prediction in digital economies. Her work bridges theoretical machine learning with practical societal impact, developing tools for safe, reliable, and equitable AI ecosystems. Her educational background includes a PhD from ETH Zurich in collaboration with IBM Research, followed by an SNSF postdoctoral fellowship at UC Berkeley hosted by Moritz Hardt. She was previously a group leader at the Max Planck Institute for Intelligent Systems before joining the ELLIS Institute. Mendler-Dünner's research spans several interconnected themes including performative prediction (where predictions change the behavior they aim to predict), algorithmic collective action (how participants can steer AI systems toward common goals), and the role of LLMs in social science research. Her work combines theoretical foundations with practical implementations, addressing challenges in interactive machine learning, optimization in dynamic environments, and context-specific evaluation of AI systems. She particularly examines how algorithmic predictions mediate services and platforms at societal scale, exploring concepts of economic power in digital markets. Her publication record shows a clear evolution from system-aware machine learning algorithms (including foundational work on IBM Snap ML) toward increasingly sociotechnical questions at the intersection of machine learning, economics, and policy. Recent work focuses on measuring performative power in digital economies, evaluating LLMs as risk scores, and developing frameworks for algorithmic collective action in recommender systems and labor markets. Among her notable recognitions are the ETH Medal for her dissertation, the IBM Research Division Award, the Fritz Kutter Award, and the IBM Eminence and Excellence Award. She is an ELLIS Scholar, a fellow of the Elisabeth-Schiemann-Kolleg, and affiliated with several prestigious research programs including the International Max Planck Research School for Intelligent Systems and the Max Planck ETH Center for Learning Systems. ETH Medal (dissertation award) IBM Research Division Award Fritz Kutter Award IBM Eminence and Excellence Award SNSF Early Postdoc Mobility Fellowship Mendler-Dünner actively mentors the next generation of researchers, advising PhD student Patrik Wolf and supervising research interns including Joachim Baumann, Haiqing Zhu, and Anna Badalyan, as well as Master's student Dorothee Sigg. She serves as core faculty for the International Max Planck Research School and associated faculty for the Max Planck ETH Center for Learning Systems. Her group has secured significant research funding through fellowships and institutional support, enabling work on projects like Powermeter (measuring search engine influence) and Snap ML (resource-efficient machine learning library with over 1 million PyPI downloads). She leads the Algorithms and Society research group, which examines machine learning as part of broader sociotechnical ecosystems. The group explores human-population interactions with algorithmic systems and incorporates these insights into learning system fundamentals. Current projects include investigating economic incentives in digital platforms, developing tools for systematic LLM evaluation in social science contexts, and creating frameworks for collective action in algorithmic systems. Mendler-Dünner also co-organizes the Algorithmic Collective Action workshop at NeurIPS 2025, demonstrating her leadership in emerging research directions at the AI-society interface.
Tapio Schneider is the Theodore Y. Wu Professor of Environmental Science and Engineering at the California Institute of Technology. His research focuses on atmospheric dynamics across Earth and other planets, climate modeling innovations, and geophysical turbulence analysis. He contributes to the Climate Modeling Alliance (CliMA) and develops advanced computational tools for climate prediction. Albert-Ludwigs-Universität Freiburg (Vordiplom, 1993) Princeton University (M.Sc. 1997, Ph.D. 2001) University of Washington, Seattle (Visiting Graduate Student, 1994-1995) His research spans climate dynamics , atmospheric turbulence , and AI-enhanced climate modeling , addressing challenges in cloud dynamics, extreme weather patterns, and planetary climate systems. Current work emphasizes hybrid machine learning-physical models and computational acceleration for high-resolution simulations. Recent publications highlight trends in AI integration for climate science, with applications in hydrology , cloud microphysics , ocean circulation , snowpack modeling , and climate tipping points . His team develops open-source tools like ClimateMachine for GPU-accelerated simulations. Scientific Recognition: Fellow, American Geophysical Union (2022) Rosenstiel Award (2019) World Economic Forum Young Scientist (2012) David and Lucile Packard Fellow (2005-2010) Alfred P. Sloan Research Fellow (2004-2006) Tapio leads climate dynamics research at Caltech, directs the Linde Center for Global Environmental Science (2011-2012), and serves as Editor for the Journal of Advances in Modeling Earth Systems . His group collaborates with NASA Jet Propulsion Laboratory (2016-2024) and Google Research (2022-present).
Virginia Smith is the Leonardo Associate Professor of Machine Learning at Carnegie Mellon University (CMU), with a courtesy appointment in the Department of Electrical and Computer Engineering. She holds a Ph.D. from UC Berkeley and undergraduate degrees from the University of Virginia. Her research focuses on developing efficient, privacy-preserving machine learning systems, particularly in federated learning, distributed optimization, and resource-constrained settings. Key areas include federated learning frameworks like CoCoA and Ditto, unlearning algorithms, and privacy amplification techniques. Smith's work bridges theory and practice, addressing challenges such as data heterogeneity, fairness, and robustness. She has led initiatives like the LEAF benchmark for federated learning and has contributed to open-source tools for scalable ML systems. Her awards include the Sloan Research Fellowship, Samsung AI Researcher of the Year (2023), and an NSF CAREER Award. She is actively involved in curriculum development, teaching courses on federated learning and large-scale ML at CMU. Her research group explores cutting-edge topics like federated optimization with sparse communication, unlearning in LLMs, and secure ML pipelines. Collaborations span academia and industry, with applications in healthcare, IoT systems, and AI safety. Smith serves as Program Chair for ICML 2025 and frequently contributes to workshops on federated learning and trustworthy AI.
Timothy M. Hospedales is a Professor of Artificial Intelligence at the Institute of Perception, Action and Behaviour within the School of Informatics at the University of Edinburgh . He also serves as VP AI and Head of Samsung AI Research Centre Europe . His research focuses on efficient and robust AI , emphasizing meta-learning , lifelong transfer-learning , and domain adaptation in both probabilistic and deep learning frameworks. Applications span computer vision , vision and language , reinforcement learning for robotics , and finance . Professor at University of Edinburgh (2020–present) ELLIS Fellow (2021) Head of Samsung AI Research Europe (2020–present) Founding Director of Applied Machine Learning Lab at QMUL (2012–2016) His work includes pioneering contributions to meta-learning , few-shot learning , and self-supervised methods , with notable awards such as the Best Paper Prize at ICML AutoML 2018 and Best Student Paper at ICPR 2018 . He has co-authored 15+ recent papers on topics like Vision-Language Models , Medical AI Fairness , and Diffusion Model Optimization . He served as Program Co-Chair for BMVC 2018 and AAAI 2022 , and authored a book on Visual Adaptation in the Deep Learning Era (2022). Co-Chair, BMVC 2018 Guest Editor, IET CV Special Issue (2016) Keynote Speaker at TASK-CV Workshop (ECCV 2016) Special Issue on Fewer Labels (IEEE PAMI 2020) His leadership extends to organizing workshops like the Learning-to-Learn Workshop at ICLR 2021 , Meta-Learning Workshop at NeurIPS 2020 , and Domain Generalisation Workshop at ICLR 2023 . Current projects include Meta-Omnium (CVPR 2023) for general-purpose meta-learning and MetaAudio (ICANN 2022) for few-shot audio classification benchmarks.
David Alvarez-Melis is an Assistant Professor of Computer Science at Harvard University's John A. Paulson School of Engineering and Applied Sciences (SEAS). He leads the Data-Centric Machine Learning (DCML) group and holds affiliations with the Kempner Institute, Harvard Data Science Initiative, and the Center for Research on Computation and Society. His research focuses on making machine learning more data-efficient and trustworthy, with applications in natural and medical sciences. He also serves as a researcher at Microsoft Research New England. Affiliations: SEAS, Kempner Institute, Harvard Data Science Initiative, CRCS Education: PhD in Computer Science (MIT), MS in Mathematics (NYU Courant), BSc in Applied Mathematics (ITAM) Research Interests: Optimal Transport, dataset distillation, interpretable AI, medical imaging, robustness, and large language models. His work bridges theory and applications, emphasizing geometric and probabilistic methods. Recent Trends in Publications: Focused on advancing optimal transport for data manipulation, distributional deep equilibrium models, and repurposing LLMs for specialized domains. Key themes include synthetic dataset generation, gradient flows in probability spaces, and robust interpretability frameworks. Awards: Aramont Fellowship, Dean’s Competitive Fund, Top Reviewer awards at major conferences (ICLR, NeurIPS, ICML). Grants: Supported by the Aramont Fund and Harvard’s Dean’s Fund. His lab advises students across Harvard and MIT, with notable contributions to medical imaging, NLP, and foundational ML theory. He actively mentors interns and fosters collaborations with industry and academia.
Geoffrey Pleiss is an Assistant Professor in the Department of Statistics at the University of British Columbia's Faculty of Science. He is also a CIFAR AI Chair at the Vector Institute and an inaugural member of CAIDA's AIM-SI (AI Methods for Scientific Impact) cluster. His work bridges statistical theory, machine learning, and computational methods with applications across various scientific domains. Pleiss received his PhD from the Computer Science department at Cornell University in 2020, where he was advised by Kilian Weinberger and worked closely with Andrew Gordon Wilson. Prior to his faculty position at UBC, he was a postdoctoral researcher at Columbia University with John P. Cunningham. His research focuses on the intersection of deep learning and probabilistic modeling, particularly on developing heuristic and approximate notions of uncertainty from machine learning models. His work has significant implications for reliable and optimal decision-making in experimental design and scientific discovery. Major research thrusts include neural network uncertainty quantification, Bayesian optimization, Gaussian processes, and ensemble methods. Pleiss develops theoretical frameworks while maintaining strong connections to practical applications across scientific domains. An analysis of his recent publications reveals a strong focus on uncertainty quantification in deep learning models, with particular attention to the limitations and capabilities of ensemble methods in the era of overparameterized models. His work increasingly addresses practical challenges in Bayesian optimization for scientific discovery, especially in materials science. There's also a growing emphasis on computational efficiency in Gaussian process methods, reflecting his commitment to making advanced statistical techniques accessible for real-world applications. CIFAR AI Chair Pleiss currently advises several graduate students including Donney Fan (PhD, Computer Science), Tim G. Zhou (MSc, Computer Science), Zachary Lau (MSc, Statistics), Nathan Cantafio (BSc, Statistics), and Tristan Cinquin (Research Intern at Vector Institute). His research is supported by multiple funding sources including his CIFAR AI Chair position, which provides significant research resources for advancing machine learning methodologies with scientific impact. Pleiss co-created and maintains GPyTorch, a highly efficient and modular implementation of Gaussian processes in PyTorch designed for speed, modularity, and prototyping. He is also involved with CoLA (Compositional Linear Algebra), a library for structured linear algebra operations in JAX and PyTorch that enables fast linear algebra computations by automatically exploiting matrix structure.
Gail E. Kaiser is a Professor of Computer Science and the Director of the Programming Systems Laboratory (PSL) in the Computer Science Department at Columbia University. She has been with Columbia University since 1985, becoming a full Professor in 1998. Prof. Kaiser's research spans software engineering, program analysis, software testing, and software security, with recent focus on addressing challenges in AI/ML systems testing and security. Prof. Kaiser received her PhD in Computer Science from Carnegie Mellon University in 1985 and her ScB in Computer Science and Engineering from MIT in 1979. Her dissertation at CMU was titled "Semantics for Structure Editing Environments" under advisor Nico Habermann, and at MIT she completed "Automatic Extension of an Augmented Transition Network Grammar for Morse Code Conversations" under advisor Al Vezza. Prof. Kaiser's research interests primarily focus on software engineering following a systems building approach, with recent emphasis on static and dynamic program analysis techniques to improve software reliability and security. Since 2005, she has investigated testing "non-testable" programs, particularly in machine learning, data mining, and scientific computing applications where traditional testing oracles are insufficient. She has developed novel techniques and tools for detecting bugs and verifying repairs in complex systems. Concurrently, she has worked on collaboration environments for computational scientists, creating knowledge sharing and domain-aware environments to support scientific workflows. Prof. Kaiser's recent publications demonstrate a strong focus on the intersection of software engineering and artificial intelligence. Her work addresses critical challenges in testing AI systems, code understanding through deep learning, vulnerability detection, and educational tools for computational thinking. There's a clear evolution from traditional software engineering topics toward AI/ML applications, with particular emphasis on metamorphic testing for non-testable systems, code similarity analysis, and educational applications. Prof. Kaiser has received numerous prestigious awards throughout her career: Distinguished Journal Award (10 Years) from 18th IEEE International Conference on Software Testing, Verification and Validation (ICST), April 2025 Best Research Paper Award at 24th IEEE International Conference on Source Code Analysis & Manipulation (SCAM), October 2024 Distinguished Reviewer Awards for ASE 2024 and FSE 2024 ACM SIGSOFT Distinguished Paper Award for "CONCORD: Clone-aware Contrastive Learning for Source Code", July 2023 Best Student Paper Award at ICCE 2021 Multiple ACM SIGSOFT Distinguished Paper Awards dating back to 2014 Presidential Young Investigator in Software Engineering and Software Systems from NSF (1988-1993) Prof. Kaiser has chaired Columbia's doctoral program since 1997 and served on editorial boards including IEEE Internet Computing and as a founding associate editor of ACM Transactions on Software Engineering and Methodology. Her lab has been continuously funded by major agencies including NSF, NIH, DARPA, ONR, NASA, and numerous companies. Current grants include significant NSF funding for secure containers architecture, learning semantics of code for software assurance, and finding semantic security bugs. As Director of the Programming Systems Laboratory (PSL), Prof. Kaiser leads research in software systems, program analysis, and software testing. The lab has developed numerous tools and techniques for software reliability and security, with recent focus on challenges in AI/ML systems. Her work bridges theoretical foundations with practical applications, often resulting in deployable tools that address real-world software engineering challenges.
Ceren Budak is an Associate Professor at the University of Michigan School of Information and holds a joint appointment as Associate Professor of Electrical Engineering and Computer Science in the College of Engineering. Her work bridges computer science, statistics, and social sciences through computational social science approaches. Her educational background includes a PhD in Computer Science from the University of California, Santa Barbara (2012) and a Bachelors degree in Computer Science from Bilkent University in Turkey (2007). Prior to joining the University of Michigan faculty, she was a Postdoctoral Researcher at Microsoft Research New York. Professor Budak's research centers on computational social science, with particular emphasis on analyzing large-scale datasets to address questions with social, political, and policy implications. Her work spans several interconnected domains: News Media Production & Consumption (examining bias in news outlets and reader preferences), Social Movements & Media (using social media data to study collective action), Social Networks (understanding information diffusion processes), and Measuring and Promoting the Quality of Online Discussions (developing tools to improve online conversations). She teaches SI 608 (Networks) and SI 618 (Data Manipulation and Analysis) at the School of Information. Her publication record demonstrates consistent contributions to understanding how online information ecosystems operate, with recent work focusing on AI-human collaboration, misinformation dynamics, social movement framing, and the application of computational methods to political communication. Her research shows a clear trajectory from foundational work on social network diffusion to increasingly sophisticated analyses of contemporary information challenges. Among her service activities, she has served as Registration chair for COSN (ACM Conference on Online Social Networks) 2015 and as Program Committee Member for numerous prestigious conferences including WWW, ICWSM, WebSci, AAAI, and others. She has also been involved in organizing the MSR NYC Data Science Seminar Series and instructing the Microsoft Research Data Science Summer School.
Kimon Fountoulakis is an Associate Professor at the University of Waterloo. His research focuses on Machine Learning on Graphs and Numerical Optimization, with a strong emphasis on algorithmic methods for graph-structured data. He holds a Ph.D. from The University of Edinburgh (2015), an M.Sc. from The University of Edinburgh (2010), and a B.Sc. from Athens University of Economics and Business (2009). His work spans theoretical foundations and practical applications in graph algorithms, optimization, and machine learning. Research interests include graph neural networks, local graph clustering algorithms, and algorithmic reasoning. His contributions address challenges in graph representation learning, message-passing architectures, and scalable optimization methods. Notable themes in his publications include improving counting abilities of vision-language models, analyzing graph convolutions, and developing flow-based clustering techniques with statistical guarantees. His work often bridges theory and practice, with applications in network analysis, pandemic containment strategies, and high-performance computing. While no specific grants or awards are listed, his research demonstrates significant contributions to graph-based machine learning and optimization. He maintains a research group at the University of Waterloo, with a focus on developing open-source tools and frameworks for graph algorithms. His lab’s work emphasizes local graph clustering methods and their scalability in real-world networks.
