Skip to main content

You're Invited

We invite you to join us for our first ever Research Computing Symposium, a full-day event showcasing research enabled by advanced computing at the University of Washington. Connect with and hear from researchers, campus and industry partners, and UWIT Research Computing experts through lightning talks, discussions, and networking sessions. The symposium will highlight current work, emerging capabilities, and the future of research computing across on-premises, cloud, and other research infrastructure.

Agenda

Main sessions are in HUB 334; check-in, breaks, lunch, and the reception are in HUB 332. During the two Research Showcase blocks, three sessions run at the same time: the main track in HUB 334 and partner tracks in HUB 238 and HUB 307.

9 AM
9:30 AM
10 AM
10:30 AM
11 AM
11:30 AM
12 PM
12:30 PM
1 PM
1:30 PM
2 PM
2:30 PM
3 PM
3:30 PM
4 PM
4:30 PM
5 PM
HUB 332
Check-in9 AM – 9:30 AM
HUB 334
Opening Remarks9:30 AM – 9:45 AM
HUB 334
Research Highlight — Paul AtkinsTraining a Deep-Learning Model to Decipher Classical Japanese Calligraphy9:45 AM – 10:15 AM
HUB 332
Break10:15 AM – 10:30 AM
HUB 334
Xu LuBanking on Inattention10:30 AM – 10:45 AM
Dylan WeissenkampeMaking a Small Study Count: Computational Robustness for Multi-Omic Human Kidney Biopsies10:45 AM – 11 AM
Aodong LiuSimulating Nuclear Quantum Effects at Scale with High-Performance Computing11 AM – 11:15 AM
Lakshitha PremathilakeOceanographic Modeling Research at the Salish Sea Modeling Center11:15 AM – 11:30 AM
Bojun ChenCARDIAC-FM: A multimodal foundation model integrating electrocardiography and cardiac MRI for cardiovascular risk prediction11:30 AM – 11:45 AM
William NobleA generalizable Hi-C foundation model for chromatin architecture, single-cell and multiomics analysis across species11:45 AM – 12 PM
HUB 238
GCP and Ai2 TrackFrom Earth Observation to AI Agents: Exploring Google's AI and Data Tools10:30 AM – 12 PMPartner track — GCP and Ai2
HUB 307
UW Libraries TrackYour Job Ran, You Got Your Results. Now What?10:30 AM – 12 PMPartner track — UW Libraries
HUB 332
Lunch12 PM – 1 PM
HUB 334
HUB 334
Research Computing at UW Panel with UWIT, eScience, and RCCResearch Computing at UW Panel with UWIT, eScience, and RCC1:30 PM – 2 PM
HUB 334
Research Highlight — Patrick BoyleVirtual Hearts and AI for Better Cardiovascular Care2 PM – 2:30 PM
HUB 332
Break2:30 PM – 2:45 PM
HUB 334
Ben WilliamsResolving Nearby Galaxies into Individual Stars2:45 PM – 3 PM
Kate HewettLiveOcean: 3-Day Coastal Ocean Forecasting in the Pacific Northwest3 PM – 3:15 PM
Yueyao Fan & Thomas HuangMachine Learning-Accelerated Atomistic Simulations of Moiré Quantum Materials at Scale3:15 PM – 3:30 PM
Emerson TurnerFull-Chip Electromagnetic Simulation for Quantum Device Design3:30 PM – 3:45 PM
Joseph LeeDesigning Protein Structures from Natural Language3:45 PM – 4 PM
HUB 238
Microsoft Azure TrackThe Next Era of Scientific Discovery2:45 PM – 4 PMPartner track — Microsoft Azure
HUB 307
AWS TrackRapidly turn ideas into working solutions with Kiro — free for UW students!2:45 PM – 4 PMPartner track — Amazon Web Services
HUB 334
Closing Remarks4 PM – 4:10 PM
HUB 332
Networking Reception4:10 PM – 5 PM
General / ProgramResearch HighlightPartner & IndustryResearch ShowcaseBreakReception

Location

HUB 332, 334, 238 & 307, University of Washington. Open in Google Maps →

Plan Your Visit:

HUB directions, parking guidance, floor plans, and FAQs.

Talks

Research Highlights

Paul S. Atkins — Asian Languages and LiteratureTraining a Deep-Learning Model to Decipher Classical Japanese Calligraphy

Premodern Japanese manuscripts were written in a cursive hand called kuzushiji, literally "collapsed characters," which few readers today can decipher. The phonetic symbols of these manuscripts, called hentaigana, are allographic, meaning that a single sound could be written with any of several forms, each a cursive version of a Chinese character known as its jibo. Many of the texts have been transcribed and printed, but those editions normally render each hentaigana as a modern kana and so discard the jibo. A scribe's choice among jibo is a feature of orthography that differs from scribe to scribe, so statistical analysis of jibo data can help answer who inscribed a manuscript, a question that has traditionally required extensive manual work. Our project applies computational paleography to this problem, training a deep-learning model to read the handwriting down to the jibo so that the analysis can be carried out at scale. The model was first trained in 2025 on a single GPU on Amazon Web Services (AWS), with cloud credits from AWS and the UW eScience Institute. The next phase is a collaboration with Tohoku University to develop a new model for retraining on H200 GPUs on Tillicum.

Computing resources: Tillicum, Cloud Computing - AWS

Abhishek Gupta — Computer Science & EngineeringTBA

TBA

Computing resources: Hyak Klone, Tillicum

Patrick Boyle — Bioengineering; Institute for Stem Cell and Regenerative Medicine; Center for Cardiovascular Biology; eScience InstituteVirtual Hearts and AI for Better Cardiovascular Care

In the Cardiac Systems Simulation Lab, we use computational physiology and machine learning to understand cardiovascular disease and develop more personalized approaches to care. Our work centers on virtual hearts — computer models built from medical images that simulate how electrical signals travel through healthy and diseased tissue. These models help us investigate why abnormal heart rhythms arise, explore connections between changes in heart structure and stroke risk, and evaluate potential treatments. We also use simulations to study emerging therapies, including why heart tissue grown from stem cells can cause rhythm disturbances. Research computing supports this work by enabling detailed, three-dimensional simulations across many patients, physiological conditions, and treatment scenarios. Alongside these simulations, we use machine learning and AI to find patterns in clinical data, including electrocardiograms and cardiac imaging, with an emphasis on interpretable methods that support earlier identification of cardiovascular risk.

Computing resources: Hyak Klone

Research Showcase — Morning

Xu Lu — Foster School of BusinessBanking on Inattention

Banking on Inattention studies how depositor inattention shapes bank deposit pricing and monetary policy transmission. Using transaction-level data from millions of U.S. depositors, we measure inattention from how long unscheduled cash inflows remain in low-interest accounts and show that banks with more inattentive depositors pay lower rates and experience more stable funding. Research computing was essential for processing billions of account-level observations, reconstructing household cash flows, and running the analysis at scale. In the lightning talk, I will highlight how large-scale research computing makes it possible to measure a traditionally difficult-to-observe economic friction — attention — and connect it to bank behavior and monetary policy.

Computing resources: Cloud Computing - AWS

Dylan Weissenkampe — Diabetes InstituteMaking a Small Study Count: Computational Robustness for Multi-Omic Human Kidney Biopsies

When a study has only a dozen participants, the hardest question is whether to believe the results. We profiled kidney biopsies from 11 people with type 2 diabetes measuring ~7,800 proteins and ~18,000 phosphorylation sites in each. Research computing let us test which signals were real: we re-ran the entire pipeline across 30 independent random seeds, built permutation-based null models over a network of ~7,800 proteins, and checked findings against matched single-cell RNA-seq data and public kidney atlases of hundreds of thousands of cells. Kopah, UW-IT's S3-compatible research storage, held the large single-cell objects and analysis outputs. Two lessons emerged: repeating an analysis many times separated robust signals from fragile ones, and a network-propagation filter only became a calibrated test after adding a permutation null.

Computing resources: Hyak Klone

Aodong Liu — ChemistrySimulating Nuclear Quantum Effects at Scale with High-Performance Computing

Nuclear quantum effects such as zero-point motion and proton delocalization play a central role in hydrogen-bonded networks and proton-transfer chemistry, yet most molecular simulations treat atomic nuclei as classical point charges. The nuclear–electronic orbital (NEO) framework provides a first-principles approach for capturing these effects. Using Hyak's Klone cluster, we developed a distributed-memory Cholesky decomposition algorithm that enabled a NEO calculation on a water cluster with 500 quantum protons — the largest NEO calculation reported to date. We have also developed GPU-accelerated NEO methods on Tillicum to further accelerate the most computationally demanding steps, enabling simulations of coupled electron–nuclear dynamics in real time.

Computing resources: Hyak Klone, Tillicum

Lakshitha (Laki) Premathilake — Salish Sea Modeling Center — UW / Pacific Northwest National LaboratoryOceanographic Modeling Research at the Salish Sea Modeling Center

The Salish Sea Modeling Center (SSMC), operating under the Center for Urban Waters at UW Tacoma, is a coastal modeling research laboratory dedicated to advancing scientific understanding of Puget Sound and the Salish Sea through hydrodynamics, biogeochemistry, and ecosystem modeling. SSMC leverages UW's Hyak supercomputing system for high-resolution model simulations and data dissemination. The real-world impact of this work spans environmental management, navigation, maritime activities, and regional stewardship — from ecosystem restoration to water quality management, marine life health, and the Salish Sea Columbia River Operational Forecast System (SSCOFS), which delivers real-time and predictive data for safe navigation and search-and-rescue operations.

Computing resources: Hyak Klone

Bojun Chen — BiostatisticsCARDIAC-FM: A multimodal foundation model integrating electrocardiography and cardiac MRI for cardiovascular risk prediction

Atrial fibrillation and heart failure impose substantial health burdens worldwide, yet accurate and generalizable risk prediction remains challenging. We developed CARDIAC-FM, a multimodal foundation model that integrates 12-lead ECG and cardiac MRI through self-supervised representation learning and cross-modal contrastive alignment. Developed on 57,609 paired ECG–cardiac MRI samples from UK Biobank, CARDIAC-FM improved prediction of incident atrial fibrillation and heart failure over contemporary ECG AI models and generalized zero-shot to two external cohorts. The learned representation transferred to continuous cardiac MRI phenotype prediction, time-to-event modeling, and additional cardiovascular outcomes with limited fine-tuning, including myocardial infarction, ischemic stroke, cardiovascular death, and all-cause mortality.

Computing resources: Tillicum

William Noble — Department of Genome SciencesA generalizable Hi-C foundation model for chromatin architecture, single-cell and multiomics analysis across species

Nuclear DNA is organized into a three-dimensional structure that impacts critical cellular processes. We have developed HiCFoundation, a foundation model trained on massive Hi-C data for integrative analysis linking chromatin structure to downstream regulatory function. The model achieves state-of-the-art performance and generalizability across species on various 3D genome analysis tasks, including reproducibility analysis, resolution enhancement, and loop detection. Additionally, HiCFoundation can predict various epigenomic activities from Hi-C to reveal how 3D structure links to regulatory function, and can adapt to single-cell Hi-C data.

Computing resources: Cloud Computing - AWS, Cloud Computing - Azure

Research Showcase — Afternoon

Ben Williams — AstronomyResolving Nearby Galaxies into Individual Stars

We have built automated pipelines to measure photometry for individual stars from high-resolution space-based imaging of nearby galaxies using both the Hubble Space Telescope and the James Webb Space Telescope. Using UW-IT and other cloud-based resources, we have cataloged hundreds of millions of stars across many galaxies in the nearby universe. The large imaging datasets and the number of stars involved require a large amount of parallel computing. I will discuss some of the computing challenges, as well as how these stars can be treated as a type of fossil record that encodes the formation and evolution of galaxies.

Computing resources: Hyak Klone, Cloud Computing - AWS

Kate Hewett — UW Applied Physics LaboratoryLiveOcean: 3-Day Coastal Ocean Forecasting in the Pacific Northwest

LiveOcean is a computer modeling system developed by the UW Coastal Modeling Group that makes 3-day forecasts of water properties in the Salish Sea and coastal waters of the Pacific Northwest. Leveraging UW Hyak Klone, LiveOcean delivers complex forecasts of currents, temperature, salinity, oxygen levels, carbon chemistry, and harmful algal blooms. Beyond forecasting, LiveOcean is run over an extended historical period (2013–present) to produce long hindcast data, allowing scientists to reconstruct past events, validate model accuracy, and explore data-driven solutions to ecological challenges. Model output is made publicly available to facilitate communication between resource users, managers, and scientists.

Computing resources: Hyak Klone, Kopah

Yueyao Fan & Thomas Huang — Materials Science and EngineeringMachine Learning-Accelerated Atomistic Simulations of Moiré Quantum Materials at Scale

Moiré quantum materials exhibit exotic electronic phenomena highly sensitive to atomic structure. We present machine learning-accelerated approaches that bridge first-principles accuracy and large-scale atomistic simulations, enabling the exploration of structure-property relationships in twisted transition metal dichalcogenide (TMD) systems. We introduce a physics-informed training-data generation strategy to develop a bespoke machine-learning interatomic potential transferable across layer numbers, stacking configurations, and twist angles. Applied to twisted multilayer MoTe₂, the model reveals layerwise structural and electronic stratification, giving rise to tunable topological states and nonlinear polarization. We further develop MoiréMLIP, a generalizable framework enabling simulations of up to one million atoms on a single GPU, revealing emergent polarization networks spanning tens of nanometers in complex moiré-of-moiré structures.

Computing resources: Hyak Klone, Tillicum

Emerson Turner — Materials Science and EngineeringFull-Chip Electromagnetic Simulation for Quantum Device Design

Superconducting circuits are a leading platform in the global effort to build large-scale quantum computers. Developing these devices requires extensive fabrication processes followed by characterization in dilution refrigerators. Electromagnetic (EM) simulation validates designs prior to fabrication; however, micrometer-scale features on millimeter-scale chips make full-device models challenging on conventional workstations. On Hyak, our group runs Palace (PArallel LArge-scale Computational Electromagnetics), an open-source solver Amazon Web Services built for quantum hardware design. Full-wave 3D EM modeling and finite-element analysis of small features enable accurate extraction of eigenmodes, field distributions, and quality factors from entire multicomponent chips, a task previous tools made impractical. Palace’s results are consistent with measurements of fabricated devices, saving time and money while improving our understanding and the overall success of our chips.

Computing resources: Hyak Klone

Joseph Lee — Computer Science and EngineeringDesigning Protein Structures from Natural Language

A central challenge in de novo protein design is generating proteins with desired structural and functional properties. We develop Prompteina, the first all-atom structure generation model that supports design objectives specified in natural language, analogous to text-to-image generation. We construct Prompteina-Data, a paired text–protein structure dataset of 1.8M structures and 50M+ captions describing structural and functional properties. Prompteina outperforms baselines on 15 of 18 structural properties and improves interface confidence by 14% over the strongest baseline in binder design. We experimentally validated designs against GM2A, a target with no reported de novo binders, identifying nine binders with the strongest at an apparent affinity of 107 nM. UW research computing resources including Hyak and Tillicum were essential in dataset generation and in training and evaluating our models.

Computing resources: Hyak Klone, Tillicum

Vendor and Partner Tracks

During the two Research Showcase blocks, partner sessions run in parallel in HUB 238 and HUB 307. These sessions are led by industry partners and campus organizations.

GCP and Ai2From Earth Observation to AI Agents: Exploring Google's AI and Data Tools

This track brings together Google Cloud and the Allen Institute for AI (Ai2) for three back-to-back sessions on how modern AI tools are reshaping research workflows. The session opens with an exploration of cutting-edge AI for scientific discovery, continues with a practical look at Google's data and AI infrastructure, and closes with Ai2's perspective on open, accessible AI for the research community.

  • AlphaEvolve (30 min) — Skander Hannachi, Google: An overview of AlphaEvolve, Google's evolutionary coding agent, and how AI-driven code generation can accelerate scientific discovery and tackle complex computational challenges.
  • Antigravity & BigQuery (30 min) — Nick Nelson, Google: A look at how researchers can use Antigravity and BigQuery to build AI-powered tools and work with large datasets — from spatial data exploration to conversational analytics and custom agent development using the Agent Development Kit (ADK).
  • Ai2 Session (30 min) — Allen Institute for AI: Ai2 shares their work on open AI research and how their tools and models support the broader scientific community.
UW LibrariesYour Job Ran, You Got Your Results. Now What?

If you want your research to have an impact, it's not enough to do good work; you've got to share it. But you've got terabytes of output, a journal asking for your data and code, a funder with its own rules, and no money for publishing. UW librarians are here to help you understand the research communications landscape and think through your options. Bring your thorniest publishing and data sharing questions, and together we'll work through what you can share, what you should share, where it might go, and who owns what.

Research Computing at UWResearch Computing at UW Panel with UWIT, eScience, and RCC

Discover how UW supports research through computing resources, training, and student programs. Representatives from UWIT Research Computing, the eScience Institute, and the Research Computing Club (RCC) will share an overview of their services, training opportunities, and student support programs. The session will also introduce two new research computing services coming in 2027. Join us to learn about the resources available to support your research and ask questions during a dedicated Q&A.

Microsoft AzureThe Next Era of Scientific Discovery

How cloud, HPC, and AI are changing what's possible in research, and what that means for your institution.

  • Welcome & Framing: The Next Era of Scientific Discovery. How cloud, HPC, and AI are changing what's possible in research, and what that means for your institution.
  • Researcher Spotlight: Lessons from the Field: Researchers talk candidly about what worked, what didn't, and the results they measured.
  • Introducing Microsoft Discovery: A first look at Microsoft's agentic AI platform. It speeds up the full research lifecycle, from hypothesis to insight, and includes built-in data collaboration.
  • Live Demo: Azure HPC + Azure AI Foundry in Action: A hands-on walkthrough of how researchers can scale simulations, build AI models, and get to discovery faster.
  • Open Q&A & Next Steps: Bring your questions. We'll also cover how to get started, including pilots, credits, MACC agreement, and technical support.
Amazon Web ServicesRapidly turn ideas into working solutions with Kiro — free for UW students!

AWS has made available free access for UW students to Kiro, an agentic, spec-driven IDE that helps you quickly set up research projects, pipelines, and tools, even if you aren’t a coder. Simply describe what you need in plain language. Kiro turns that into requirements and a design plan, then writes, tests, and documents the code with you. In this session we'll give a quick overview of Kiro, walk through some demonstrations, and then show you how UW students can get it for free. Drop in and learn how you can get to research outcomes faster. More information at https://kiro.dev/students where students can sign up free with their @uw.edu email address - no credit card needed.

Campus Partners & Sponsors

We are grateful to the campus partners and sponsors who make this event possible.

Event Sponsors

Dell Technologies
Redesign

Industry Participants

Amazon Web Services
Google Cloud
Microsoft Azure

Campus Partners

UW Office of Research
eScience Institute
UW Libraries

Related Event

Accelerating Research Day

A half-day engagement on Thursday, November 12, 2026 hosted by Dell Technologies and NVIDIA. Explore the transformative power of AI through strategy sessions and hands-on labs, learn from experts, and network with peers. Lunch will be provided.

Learn more about Accelerating Research Day →