ÌìÃÀ´«Ã½

What’s ahead for research computing at Imperial? In this opening session, Andrew Richards introduces the Research Computing Showcase Day 2026 and sets out the RCS roadmap, highlighting priorities, upcoming initiatives, infrastructure direction, and opportunities for researchers, research software engineers, and HPC users.

Whether you work in AI, simulation, data science, HPC, cloud, or research infrastructure, this talk gives essential context for the day and for the future of institutional research computing support.

Research computing doesn’t stop at institutional boundaries. In this session, Michael Bearpark explores activity beyond Imperial, including Network+ initiatives, RTP projects, collaborations, and wider research technology partnerships that help expand capability, expertise, and impact.

This is a useful overview for researchers interested in cross-institution collaboration, digital research infrastructure, funding opportunities, and strategic partnerships in the UK research computing ecosystem.

In this launch session, Dugan Witherick introduces HX2 (HEX-HTC) and HX3 (HEX-AI), Imperial’s newest high-performance computing platforms. Learn what these new systems are designed for, who they support, and how they will help accelerate research across AI, simulation, modelling, data analysis, and computational science.

What is happening across the UK’s digital research infrastructure landscape? In this presentation, Ben Yarnell shares an update on the UKRI and DSIT vision for the UK Research & Innovation Compute Landscape and the wider transformation programme.

The talk covers the principles of UKRI’s digital research infrastructure strategy, progress against compute transformation objectives, and recent investments in high-performance computing. It also explains how new infrastructure will function and fit into the wider national landscape.

This session is essential viewing for anyone following UK HPC strategy, research infrastructure policy, compute investment, and the future of large-scale scientific computing in the UK.

How can researchers identify meaningful gene expression patterns across multiple studies while accounting for differences between populations and datasets? In this talk, Margarita Grushhanina presents a computational statistical framework for joint analysis of multi-study gene expression data.

The method combines structured pattern discovery with regression on individual-level covariates, while clustering genes and individuals with similar expression patterns. It explicitly separates study-specific variation from shared biological signal, helping researchers distinguish real biology from technical artefacts.

The presentation also covers a sparsity-inducing Bayesian formulation, variational inference, and how high-performance computing enabled large-scale simulations, hyperparameter tuning, and robust evaluation on high-dimensional genomic data.

A great session for anyone interested in computational biology, bioinformatics, Bayesian statistics, machine learning, gene expression analysis, and scalable genomics.

What does it take to support research outputs effectively at scale? In this session, Lalitha Kambhammmettu introduces Helix and its role in improving how research outputs are supported, organised, and surfaced.

This talk will be especially relevant to researchers, support staff, and research technology professionals interested in research data, software outputs, visibility, discoverability, and infrastructure for modern scholarship.

In this session, Diego Alonso Alvarez introduces a new structure and new initiatives within the Research Software Engineering team and research support landscape. Learn what is changing, why it matters, and how these developments aim to improve support for researchers, software, infrastructure, and innovation.

This is a useful update for anyone who wants a clearer view of service evolution, strategic direction, and future opportunities in research computing.

Can high-order CFD solvers scale efficiently from a single GPU to many? In this presentation, Henrik Wustenburg shares early results from a major redesign of Nektar++, the open-source spectral/hp element framework used for solving partial differential equations in computational fluid dynamics and related applications.

The talk explains how the new architecture targets heterogeneous HPC systems using backends such as CUDA, HIP, SYCL, and SIMD vectorisation, enabling performance portability across diverse hardware platforms. It also presents early performance and scaling results for global operators and initial solver-level tests for unsteady diffusion and advection equations.

This session is ideal for viewers working in CFD, PDE solvers, GPU acceleration, HPC performance portability, scientific software, and exascale-ready simulation.

Aeroelastic effects in turbomachinery are critical to the safe and efficient operation of modern gas turbines. In this talk, Connor McCleod presents high-fidelity aeroelastic simulations for turbomachinery applications using HPC.

The presentation focuses on AU3D, an aeroelastic solver developed at ÌìÃÀ´«Ã½ and used by Rolls-Royce in product design and certification. It explains how the solver captures complex fluid-structure interaction, coupling unsteady aerodynamic loads with structural response to model blade displacements and mesh updates over time.

The talk also shows why high-performance computing is essential for demanding simulations of non-synchronous vibration, unsteady flow, and aeroelastic instability, and how these methods support certification and component development in aerospace engineering.

Perfect for audiences interested in aeroelasticity, CFD, fluid-structure interaction, turbomachinery, gas turbines, industrial simulation, and HPC engineering.

Research software is a vital research output, but it is often difficult to showcase, track, and evaluate. In this presentation, Diego Alonso Alvarez introduces the Imperial Research Software Directory, a platform designed to help the university tell the story of its software and better understand its impact.

Built using technology from the Netherlands eScience Centre, the directory can scrape metadata from open-source repositories such as GitHub, GitLab, and Bitbucket, track references and citations from sources including Crossref, DataCite, OpenAlex, and Software Heritage, and connect collaborators and maintainers through ORCID.

The talk explains how the directory improves software discoverability, impact tracking, research visibility, and institutional understanding of software as a research output.

Ideal for anyone interested in research software engineering, software citation, open source, scholarly infrastructure, and research impact.

Space weather can disrupt power systems, satellites, communications, and critical infrastructure. In this talk, Adrian LaMoury explains how high-performance computing is being used to improve space weather forecasting by modelling Earth’s magnetosphere and its interaction with the solar wind.

The presentation introduces the Gorgon codebase, originally developed at Imperial for laboratory plasma experiments and later adapted for planetary magnetospheres using a magnetohydrodynamic approach. Recent developments with RCS have made Gorgon suitable for operational forecasting, using live satellite measurements as inputs and running parallel simulations on CX3 Phase 2 faster than real time.

The talk also covers cloud export workflows, APIs, and adoption by organisations including the Met Office and the European Space Agency, showing how advanced simulation can help protect vital infrastructure.

A must-watch for anyone interested in space weather, plasma physics, magnetospheres, HPC, forecasting, and scientific modelling.

How can research computing become more sustainable? In this session, Chris Cave-Ayland highlights sustainability initiatives at Imperial RCS, exploring practical ways to reduce environmental impact while maintaining performance and supporting cutting-edge research.

This talk will interest researchers and technical teams focused on green computing, energy-efficient HPC, sustainable infrastructure, responsible AI, and operational best practice.

Strong research computing depends on strong engagement. In this session, Emily Lumley explores research engagement across users, services, and platforms, showing how effective communication and collaboration help researchers get more value from available expertise and infrastructure.

This talk is relevant for anyone interested in community building, user support, research enablement, digital skills, and improving uptake of research computing services.

With demand for GPUs soaring for AI and simulation workloads, can modern CPUs handle some of the pressure? In this talk, Max Gamill explores whether modern CPUs can provide a practical alternative for repetitive, compute-intensive tasks on academic HPC clusters, helping reduce dependence on scarce GPU resources.

The session examines CPU and GPU usage on Imperial’s CX3 and HX1 systems to identify common GPU workflows and performance bottlenecks affecting researchers. It also benchmarks modern CPU hardware using representative AI workloads such as AlphaFold, with a focus on Intel Granite Rapids, Advanced Matrix Extensions (AMX), and MRDIMM high-bandwidth memory.

This is a valuable talk for anyone interested in HPC performance optimisation, AI infrastructure, CPU vs GPU trade-offs, cluster efficiency, and workload scheduling.

Biomedical evidence on diet-related non-communicable diseases is spread across large volumes of heterogeneous HTML and XML literature, creating major barriers for large-scale text mining and data reuse. In this talk, Joram Posma presents Auto-CORPus and CoDiet, an end-to-end pipeline for converting unstructured diet-NCD publications into a standard, machine-interpretable format for natural language processing and deep learning.

The session explains how full texts were downloaded and standardised into a common BioC JSON representation, and how the RSE team helped professionalise Auto-CORPus into a reusable command-line tool. It also introduces TeamTat, a custom Azure-based collaborative annotation platform used by more than 40 annotators to label 500 full-text documents.

The result is a major new biomedical gold-standard corpus for named entity recognition, relation extraction, knowledge graph creation, and LLM-ready literature mining, designed to speed up AI-assisted evidence synthesis while helping future-proof evaluation against data leakage.