At the start of September, several members of The University of Manchester’s Research Software Engineering (RSE) team hopped on the train across the Pennines to attend the 10th annual RSE Conference (RSECon26), hosted at The Wave at The University of Sheffield. Celebrating a decade of RSECon was already a brilliant milestone, but it was an extra special trip for both of us. Having just kicked off our Year in Industry placements with the Manchester RSE team back in July, we were lucky enough to tag along for our very first taste of an academic conference.

RSECon26 is all about bringing together software engineers, academics, and partners across the community to talk about what is working, what is tricky, and where computational science is heading next. This year, the conference rallied around two central themes: RSEs as part of the research journey, and Enhancing credit and reproducibility: research software quality, performance, and evidence. Across three jam-packed days featuring 53 talks, nine workshops, eight Birds of a Feather (BoF) sessions, and 24 poster pitches, we got a proper ringside seat to the whole ecosystem.

Manchester Taking the Stage

Naturally, we spent a fair chunk of time cheering on our own Manchester colleagues, who seemed to pop up everywhere across the programme.

Andrew Rowley treated the room to a fascinating lightning talk on Manchester’s SpiNNaker machine: a neuromorphic supercomputer boasting over one million ARM cores, purpose-built to simulate real-time spiking neural networks. Seeing the architectural schematics and physical boards on the big screen was a great reminder of just how ambitious our home institution's hardware engineering can be. Andrew also took part in the Coding Confessions workshop, where engineers openly shared past mistakes and war stories to normalise the trial-and-error side of writing software.

Sarah Jaffa co-led an energetic Birds of a Feather session, RSEs as trainers: the who, why and how of providing digital research skills training, joining forces with peers from the National Oceanography Centre, UCL, and the Met Office to dig into how we formalise teaching across institutions. True to form, Sarah later stepped up for a lightning talk with a campaign everyone could get behind: "Bring back Reading Week".

Meanwhile, Adrian Harwood and Phil Bradbury teamed up with colleagues from Durham and York to introduce the DisCouRSE Network+ funded project dRTP COMPASS, a collaborative network designed to offer operational management advice, peer support, and shared mentorship for digital Research Technology Professionals across the UK.

Coding for the Real World & Deep GPU Profiling

Outside our own backyard, several sessions and practical workshops really showed how research computing tackles everything from raw hardware bottlenecks to societal challenges.

On the high-performance computing side, AMD held a hands-on workshop called From Bottleneck to Breakthrough: A Hands-On Workflow for Profiling and Optimizing HPC Applications on AMD GPUs. It tackled a headache every developer faces: writing the code is often the easy part, but figuring out where performance is actually being lost is much harder. Using a 2D shallow-water solver as a live running example, the session guided attendees through the ROCm profiling ecosystem, covering kernel analysis, roofline modelling, and multi-node scaling with access to AMD MI300A APUs.

In her Emerging Voices keynote, Yigakpoa Ikpae from Open Gov Africa asked a simple question that hit home: why is public government data so notoriously difficult to use? It turns out the answer is usually that it is scattered across random websites, buried in unreadable PDFs, or published in baffling formats. She pointed out that this gap between information merely existing and actually being usable is the exact same bridge RSEs build every single day when wrangling raw research data into accessible code.

We saw a wonderful example of that bridge in action during Evie Grace’s talk on Day 3. She walked us through a Quarto geospatial analytics toolkit created by Bristol’s Jean Golding Institute alongside local councils. The toolkit directly supplied the geographical evidence base local policymakers needed to greenlight the WEST regional policy offering free bus travel for kids. Seeing lines of Python and Quarto dashboards turn into free bus tickets for young people was proof of how tangible this work really is!

Sheffield’s team also showed off their verification plugin, which packaged over 20 multimodal AI research models into a nifty browser extension used by more than 140,000 journalists to detect manipulated media and debunk fake news on the fly.

Exploring the Stalls & Networking at The Wave

Between the talks, the atrium at The Wave was buzzing with activity. The sponsor and community exhibition stands gave us a fantastic opportunity to see the wider industry and national research network up close.

Saad spent his break time exploring the exhibition floor. Over at the AWS booth, he chatted with a current solutions architect apprentice about early-career pathways in the cloud. At the CAKE (Computational Abilities Knowledge Exchange) stand, he explored community fellowships and cross-institution visit schemes. Over at the Met Office stand, he got to inspect physical supercomputer compute blades and node reference maps up close. To top it all off, visiting the AMD exhibit led to a great conversation with an engineer who had travelled all the way from Texas to Sheffield!

Saki took the opportunity to dive into the community side, using her go-to question: "What was your journey to becoming an RSE?" She quickly discovered how diverse the community really is. People came from astrophysics, bioinformatics, digital humanities, and self-taught software backgrounds, all bringing unique perspectives to the role. It was eye-opening to see how common challenges, such as rescuing legacy code or introducing version control to non-technical teams, connect engineers across entirely different disciplines.

What We Learned & Taking It Back to Our Desks

Coming back to our day-to-day placement tasks, so much of what we heard reframed how we think about our projects. In his opening keynote on trustworthy science, Arfon Smith dropped a line that stuck with both of us: the true KPI of research software is not raw lines of code; it is the research and new knowledge it enables.

For Saad, this resonated immediately while building data ingestion pipelines for the Digital Cancer Research team: if pipelines are not reliable, tested, and verifiable, the downstream science cannot be trusted. For Saki, attending the Coding Confessions and training sessions drove home the importance of building sustainable, maintainable software from day one, rather than treating documentation and error handling as an afterthought.

We also appreciated the open culture around well-being. Between the ConveRSE sessions and the wellbeing checklists pinned up around The Wave, it was reassuring to learn that wrangling technical debt, context switching, and imposter syndrome are universal developer experiences.

Expectations vs. Reality

Before arriving, we both wondered whether an academic conference might feel slightly formal and intimidating. In reality, it was about as welcoming as you could imagine.

Saki loved realising how generous everyone was with their time and knowledge. Whether discussing how to preserve fragile legacy research or learning how different universities structure their teams, there was zero elitism, just a genuine desire to collaborate, share open-source tools, and mentor newcomers.

Saad loved seeing the balance of deep technical engineering and light-hearted fun. Between geeking out over GPU acceleration, watching Dylan Beattie rock out his comedy coding gig at the Cutlers’ Hall dinner, and spotting a custom LEGO set built for the programme chair, the whole week proved that serious research does not take itself too seriously.

Final Thoughts

Heading into the rest of our placement year, we could not be more excited to keep learning, building, and contributing back to the team!