Research Software Engineering conference 2018 #RSE18

On 3rd September the peaceful campus of Birmingham University came alive with bubbling groups of research software engineers, talking in excited tones about their latest optimisation tool and favourite python library, as the third annual conference of Research Software Engineers was started!

A real global affair, #RSE18 had 314 delegates from 12 countries. That represents a nearly 50% increase over last year’s attendance and also a 7% increase in women attending compared to 2017.

Big News!

UK RSE Association is turning into the Society of Research Software Engineering! A legal, independent, professional organisation!

The UK RSE Association has seen significant growth since its inception in 2013, to over 1000 members. The community’s growth has made the informal, volunteer run format unsustainable. The move will enable the society to hold funds, employ staff, and operate as an independent organisation to represent the interests of the RSE community. Visit the RSE website for more information and sign up to receive updates.

Keynotes

Eleanor Robson (Professor of Ancient Middle Eastern History and Head of the History Department at University College London) kicked the conference off with the first keynote about Nammu and Oracc, two digital humanities software packages. Nammu is an easy-to-use text-editor for cuneiform (a written language from c. 3500-3000 BCE) inscriptions. Oracc is a collaborative online publishing platform for cuneiform transcriptions. Eleanor and a team of RSEs from UCL are developing these tools to aid with the protection, conservation and documentation of some of the world’s oldest written language. It was exciting to hear how computational approaches can be developed in all subject matters, even ancient history!

The second keynote of day 1 was given by Andrew Fitzgibbon from Microsoft Research on “Building computer vision systems that really work”.  He spoke on the technology and software that is working in Microsoft’s HoloLens, how it grew from 2D point mapping to real world camera tracking and in-situ generation of 3D geometry. Andrew talked passionately about how doing ‘research’ and ‘engineering’ should always be done with a real-life application as the end goal. It’s how our generation can positively affect the future through the science that we do.

Later in the day I got a chance to try the Microsoft HoloLens. It was incredibly interesting to see how it mapped the environment around you in real time. It was relatively robust with flat surfaces such as floor, stairs, and walls but had trouble picking individual people out in a busy environment. It took a few minutes to get used to the ‘pinch’ controls but once I got to grips with it, manipulating the augmented environment was quite intuitive. Other than the field of view being incredibly small I can see lots of applications for this kind of tech. Excited to try the next iteration!

Day 2 was kicked off by James Howison from University of Texas at Austin talking on sustainability in scientific software. As software has become more important to the practice of research, policy makers and researchers have become concerned about a perceived lack of sustained benefit from software produced by grant funded projects. James expressed that ~20% of the cost of a code’s lifetime is in development, the rest is maintenance. The software ecosystem of a development project is complex and sustainability should be considered at every step, from proposal to publication. Academic software has a unique issue as standalone branches developed for a particular area of research are often not merged back to the master. Leaving unfinished and unsupported strands of a software package just floating.

Andreas Fidjeland from Google’s DeepMind finished the conference’s keynotes with a talk on how machine learning and AI are transforming research. DeepMind’s scientific mission is to push the boundaries of AI, developing programs that can learn to solve complex problems without needing to be taught how. Andreas talk about how AI was trained to play Go, a game with very simple rules but incredibly complex permutations and number of potential moves. When world leading Go players play with intuition, it is not a trivial task to convert their logic to physical quantities that can be evaluated.

Talks & Workshops

Between the keynotes (and networking over snacks and coffee of course!) various parallel sessions were held. These consisted of both workshops and talks and ran over the two days. The general theme of the sessions that I attended was enabling better software through guidelines and tools, the conference had a big emphasis on knowledge and skill sharing to better equip RSEs in their roles. This was echoed in the talks and especially the workshops. There were workshops focusing on best practices for programming languages such as C++ and Python, and discussing technologies such as containerisation and machine learning.

Another of the conference’s aims was to help foster a welcoming and connected community with RSEs regardless of location. This was done through a series of talks that focused on community and how to build a thriving group. Not just a community between fellow RSEs but with funders and wider research fields.

Diversity is another theme, and a workshop was held to discuss best practices for inclusive hiring – making us think about how to reach the best staff for the job, and give everyone an equal chance through the interview and selection process.

Among the many discussions that took place, the issue of recognition and reputation was brought up many times. The general consensus was that RSEs are under recognised for their contribution to academic research. It is not common place to be acknowledged for work completed on software that was used by an academic, therefore it can be difficult to build an external reputation and measure the impact your work has had. This is one of the goals of the RSE Society is to put in place a recognition scheme on-par to publication.

High Performance Computing Workshop

Aiman Shaikh and I stayed an extra day in Birmingham to attend the RSE in High Performance Computing day workshop. Approximately 10% of conference delegates stayed for this extra workshop.  Its aims were to bring together RSEs working in the HPC space.

The day was split in two, with talks from tier-1 and tier-2 HPC representatives in the morning and a discussion on HPC Champions in the afternoon. The talks covered a wide range of HPC topics, from open source benchmarking tools to developing collaborations through hackathons. After some interesting talks (and more coffee and pastries!) we were introduced to HPC Champions.

HPC Champions is the next step on from the ARCHER Champions programme. An ARCHER Champion is an individual within a regional or national centre who has the ability to support and advise users on how to access the HPC resources available to them in the UK. They can also be an RSE that works with a group of HPC users to provide support and skill sharing within that group. HPC Champions are going to have a similar description, but with more emphasis on championing excellent HPC software practices. HPC Champions is not just for RSEs within the HPC space, its aim is to be inclusive to all who work with HPC; RSEs, Supporters, Researchers, and Industry. HPC Champions is going to be linked to HPC UK.

Final Thoughts

Overall it was a great experience attending RSE18! Being surrounded by like-minded people who have gathered to openly discuss and share resources and skills was invigorating. There is a real community developing for RSEs and it is exciting to be here at the start of the RSE Society. Topics ranged from best practices to better user interfaces to programming tools for easy and faster code to cutting edge science. I particularly enjoyed discovering how what I thought was a purely scientific role can be extended to a larger range of subjects.

Tim Powell

@MrTimPowell

@ResearchSoftEng