By Gillian Law
Research software ought to be easier to use, says newly appointed 2018 RSE Fellow Jeremy Cohen.
A computer scientist by background, Cohen has spent the bulk of his career to date in the Department of Computing at Imperial College London, supporting scientists in their research. “I’ve worked in a lot of domains, offering applied computing research to support scientists in various areas” he says. This has included using high-performance computing platforms and cloud-computing infrastructure.
“Of course, in general scientists tend to have some computing knowledge, but they often have very much a domain-focused view,” Cohen says. He aims to make their codes easier to access and use. Even if the scientist could do it themselves, they may end up doing it in a more complicated or inefficient way if they’re learning as they go along, “so I aim to make life simpler and let them focus on the science, not the computing.” Good code can also help scientists to make their modelling and simulation work more accessible to their broader team. “There’s often a lot of post processing needed on a model or simulation and so what I’m trying to do is bridge the gaps, or glue together different processes, and simplify complex things. Again, we’re working with people who are very experienced and they can do this work themselves, but we can help them to do it much more effectively.” That opens up the code to a broader group of users.
“If you want to scale and increase the number of people that use your code and make it available to a wider community, you can’t continually scale your development team at the same time, with more and more people to support all those users. What I’ve been trying to do is help people in that position by providing web-based tools that make it much easier for the user base to work with the codes.”
As part of his fellowship Cohen will work with a team from Imperial College’s Aeronautics department who are involved in development of the Nektar++ spectral/hp element code. He also hopes to apply solutions to other codes within the Platform for Research in Simulation Methods (PRISM) community at Imperial. “There are a number of different teams within the PRISM platform who provide simulation software and tools for different scientific communities. I think there’s scope to apply some common ideas between those codes,” he says. Cohen will also work with the College’s Bioinformatics Data Science Group (BDSG) and the Institute for Translational Medicine and Therapeutics (ITMAT) Data Science Group. “This is a huge area. If you’re a bioinformatician many standard bioinformatics processes and tools are second nature to you, you understand the processes and what the different options are, but there are large numbers of people coming into the community who don’t have that kind of knowledge. The scientists in the domain therefore want to make the processes more generally available, even if only to simplify their own work. So we’re looking at tools to achieve that,” he says.
“It’s very much a work in progress and we see a lot of scope to do interesting work here.”
In 2015 Cohen started a research software engineering community within Imperial College, aiming to bring together software developers, researchers, academics and research managers with an interest in the development of research software. The community now has a mailing list of around 150 people, and Cohen hopes to expand this group as part of his RSE fellowship. “Imperial College is very much focused on science and engineering, and there are lots of people building software and working with it. When that grows organically you get this very siloed approach, with people in each domain doing their own stuff – but they’re all trying to learn, and they’re all using very similar techniques and technologies. What an RSE group does is help cross-pollinate between the communities,” Cohen says. “If you put some framework around the tools that you teach people and bring these people together, tremendous benefits can come of it. So I set up this community to foster that linking of people in different areas.”
A final, smaller thread to Cohen’s fellowship will be to do an economic analysis of the benefits of RSEs to an institution. “People are setting up RSE groups because they think it’s beneficial to support their local scientific communities, and they’re willing to invest to see how it works. But the actual economic benefit of doing things one way versus another has not been fully explored, so potentially that could be exciting,” he says.