Speakers and Talks - RSE South West at the University of Bath

Speaker details for talks in the Friday 31st July 2026 programme.

Introduction

Stephen Cook

Affiliation: University of Bath

Talk: Introduction to the event

James Thomas

Affiliation: Jean Golding Institute, University of Bristol

Talk: Intro to the RSE South West mailing list and website

Keynote

Ed Bennett

Affiliation: University of Swansea

Talk: SHAREing

Contributed Talks 1

Tom Burnett

Affiliation: University of Bath

Talk: An Introduction to the IMI

Summary: The Institution for Mathematical Innovation (IMI) adds value and Impact to research, creating interdisciplinary connections through the power of modelling, data science, and machine learning. We achieve this via the inclusion of Mathematical Innovation Research Associates (MIRAs) on grant submissions, through industrial and academic partnerships, and by supporting research excellence through our small grant scheme and grant enhancements.

Miquel Perello Nieto

Affiliation: University of Bristol

Talk: Launching our Mathematics Innovation Initiative

Summary: A team of academics and researchers from the University of Bristol is planning to create an Innovation Centre in Mathematics to help researchers connect with businesses, industry and government. The new team of Innovation Research Associates will bring mathematical insights into real-world projects and help organisations solve complex challenges.

Wilf Wilson

Affiliation: Met Office

Talk: SSEs in Space Weather at the Met Office

Summary: Space weather forecasting is a relatively new area at the Met Office, and has been growing in recent years, including with the addition of Scientific Software Engineers. There are now three of us, with two more currently being recruited, supporting the scientists in their work. I’ll give a quick introduction to our group.

Elizabeth Mamchits

Affiliation: STFC/DAFNI

Talk: Five Lessons from Developing and Supporting the DAFNI Research Platform

Summary: Research software platforms need to balance robust engineering with the varied requirements of researchers from different disciplines. This talk uses the Data & Analytics Facility for National Infrastructure (DAFNI) platform as a case study to explore the challenges of developing, deploying and supporting a production research platform. Through a live demonstration, I will show how researchers interact with DAFNI before discussing the engineering decisions behind the platform, the role of user engagement in shaping development, and lessons learned from supporting an active research community. The session aims to share practical insights that other Research Software Engineers can apply when developing research-facing software and infrastructure.

Lewis Sampson

Affiliation: STFC/DAFNI

Talk: Through the lens of the DAFNI, maintaining a digital research infrastructure platform with industry standard tooling

Summary: The DAFNI platform is supported by on-prem hardware and STFC hosted service, our team manages a Kubernetes cluster with GitOps deployment of APIs and other industry tooling. I will talk through the architecture and touch on the CI/CD cycle of our current platform with non-technical descriptions for the tooling involved. Then I’ll summarise with a pros/cons/recommendations that we’d like to make to others.

Contributed Talks 2

Mateus Bernado-Harrigton

Affiliation: Cardiff

Talk: Pair Programming as Peer Mentorship: Building Research Software Communities One Session at a Time

Evie Grace

Affiliation: Jean Golding Institute, University of Bristol

Talk: Plotting the route from research engineering to policy: a Quarto geospatial analytics toolkit built with and for local government

Summary: Local authorities hold a wealth of geospatial transport data that has the power to provide critical insights into transport networks in their constituencies. However, this data may not be research-ready or maximised to the full due to constrained resources. Urban Ride Research is a collaboration between the University of Bristol and the West of England Combined Authority (WECA) to tackle this gap. We developed a Quarto toolkit for transport data analysis, including a dashboard with interactive plotting that draws on both WECA data and open-source demographic data to provide analytical insights. The toolkit is designed to be flexible to meet WECA’s needs: it is written in Python, and Quarto was chosen so that R developers can extend the work in their preferred language. It also supports easy customisation of analysis depending on the policy use case. To support longevity, we are developing training and documentation, so WECA has autonomy over the code and future development. A central use case has been WECA’s Kids Go Free scheme, that delivered over 910,000 free under-16 bus journeys in summer 2025, and which the toolkit allows analysts to interrogate by comparing usage patterns across holiday and term-time periods. Beyond data science, this collaboration has opened channels between local government and researchers, through a data sharing agreement and developing reusable analysis tools. The talk will explore the role of RSEs in working with local government to understand their needs, ambitions and policy landscape, to provide tools that empower policy makers in data-driven decision making.

Emma Hogan

Affiliation: Met Office

Talk: What Happened Next? Advancing learning resources for iterative, customer-focused development of research software

Summary: At the last RSE SW meeting in February 2026, I explored the need for learning resources to support iterative, customer-focused development practices that help deliver high-quality research software that lasts. In this lightning talk, I will reflect on progress since then, share emerging insights from activities carried out over the past few months, and discuss potential next steps for developing and promoting learning resources in this space.

Elvira Alier Retamar

Affiliation: University of Bath

Talk: Hot Topics Coming to DDaT

Summary: Hot Topics Coming to Digital, Data and Technology Department at the University of Bath: Storage, HPC and Cyber Updates for Research Computing

Josh Tyler

Affiliation: Jean Golding Institute, University of Bristol

Talk: Have I Learnt Python?

Summary: The phrase “I read it better than I can code it” is one that is often said by R and Python users about the other language. Whilst they share many approaches and tools, there are some quirks that are quite fundamental. This talk will share some thoughts from an R user, whose first major Python project was to build a package from scratch, and whether that is a good way to learn. I will also showcase the final python Library and the GUI, which was built using shiny for Python.

Contributed Talks 3

Francesco Devoto

Affiliation: Met Office

Talk: Machine Learning for Visibility Range Forecasting and Classification

Summary: Autonomous vehicles (AVs) are becoming increasingly widespread, raising concerns about safety and the impact of adverse weather on their sensing and decision-making systems. By replacing the human driver with a suite of sensors and an Automated Driving System (ADS), it becomes essential to understand how these components perform under varying environmental conditions. AVs rely on multiple sensors to detect obstacles and avoid collisions; however, factors such as sunlight glare, rainfall, road spray, and water on sensors can significantly degrade performance. The ADS must be able to account for these effects and respond appropriately to ensure safe operation. In this project, I analyse meteorological data collected during an experimental campaign involving sensors used in autonomous vehicles and apply machine learning models to forecast and classify visibility ranges. The resulting models could be integrated into ADS frameworks to enhance situational awareness and improve safety under reduced-visibility conditions.

Honey Singh Chauhan

Affiliation: University of Bath

Talk: Thousand calculations - multiple options: Choosing an HPC submission strategy

Summary: Researchers often need to run hundreds, sometimes thousands, of independent calculations. The obvious question is usually, “How many CPUs should each job use?” But there is another question that matters just as much: how should all those jobs be submitted? In this study, I used Gaussian workloads to compare several common approaches, including individual jobs, Slurm arrays, throttled arrays, bundled jobs. I looked beyond the runtime of a single calculation and focused on the whole campaign: completion time, throughput, scheduler pressure etc. The aim was to find a submission strategy that is not only fast, but also practical, resilient and considerate of a shared HPC system.

Rita Rasteiro

Affiliation: University of Bristol

Talk: DEICAP: Measuring the Hidden Carbon Cost of Digital Research

Summary: Despite rapid growth in digital infrastructure, most UK universities lack the tools, standards, and data models needed to quantify and integrate digital-related emissions into decision-making. DEICAP (Digital Emissions Integration for Climate Action Planning) addresses a critical gap in sustainability across universities by improving understanding of digital carbon emissions. The project will update and extend the University of Bristol’s existing digital emissions model to reflect new digital technologies and infrastructure and apply it across partner institutions. It will also develop shared tools and reporting frameworks to improve transparency and support carbon-informed decision-making, helping institutions reduce their digital carbon footprint and accelerate progress towards Net Zero.

Dr. Harry T Mason

Affiliation: University of Bristol

Talk: Assessing the impact of COVID-19 on neurodegenerative diseases in vaccinated and unvaccinated people, using the secure OpenSAFELY platform

Summary: We used OpenSAFELY-TPP to access linked electronic health records from 18million English adults. We estimated adjusted hazard ratios (aHRs) comparing the incidence of cognitive impairment, dementia, and other conditions up to 5.5 years after COVID-19 in three cohorts (pre-vaccination; vaccinated; unvaccinated). Using OpenSAFELY allows us to access large amounts of linked secure data while also keeping our code open source. Due to the security of the platform, this transparency is allowed without violating the 5 safes principals. Incidence of most outcomes were elevated during weeks 0-4 after COVID-19 but attenuated beyond 4 weeks. Incidence of cognitive impairment remained elevated up to 4 years after COVID-19 (aHRs, pre-vaccination:1.13(1.09-1.17); vaccinated:1.15(1.11-1.19); unvaccinated:1.37(1.22-1.55)). Our findings suggest a potential longer-term impact of COVID-19 on cognitive health, particularly among individuals unvaccinated at the time of COVID-19 diagnosis.

Roberto Scipioni

Affiliation: Red Oak Consulting

Talk: EasyBuild and EESSI the Initiative for Software Usage on HPCs

Summary: A description of the EESSI Software Initiative that through the usage of files and software layers allows the usage of a big collection of software on diff infrastructures.

Ashutosh Shankarrao Londhe

Affiliation: University of Bath

Talk: EuroHPC Access for UK Researchers: Opportunities Beyond Local HPC

Summary: EuroHPC provides UK researchers with free access to some of Europe’s most powerful supercomputing systems through competitive access calls, yet many researchers and Research Software Engineers remain unaware of these opportunities. This talk provides an overview of the EuroHPC programme, introduces the available access modes and the LUMI supercomputer, and explains how researchers can apply for compute allocations. Drawing on my experience of securing multiple successful EuroHPC awards for code development, benchmarking, and GPU enablement, I will share practical insights into the application process and discuss how EuroHPC resources can complement local HPC facilities, enabling larger-scale computational research without requiring dedicated project funding.

Discussion

Ben Westhenry

Affiliation: Jean Golding Institute, University of Bristol

Discussion topic: How do you use GenAI for coding?

Summary: How do you or your group use GenAI for coding? Do you get help from a chatbot, or do you use a fully agentic workflow? In this session, I’d like us to discuss what different people do, any tools or resources they make use of, and any barriers they face.