We recently launched the Research Software Capability in Australia report, which describes the scale and distribution of software engineering or development capability within research organisations in Australia.
One organisation stood out in the report with its recognition of research software. Australia’s national science agency, CSIRO, has operated a dedicated group for addressing research software challenges across the organisation for over 8 years.
Justin Baker, who leads the Data Analytics and Visualisation group as part of CSIRO’s Scientific Computing Services, said that this is a diverse capability, with a mix of research and professional backgrounds and experience.
“Those with domain knowledge can really hit the ground running. For example, data analytics or genomics is a huge field, and if you’ve already got some background in that, the software projects progress a lot faster.”
CSIRO has a comprehensive process for deciding which projects receive research software engineering services. These are known as Scientific Computing Collaboration Projects.
“We advertise internally in CSIRO for a request for proposals every 6 months. Researchers or research groups respond by submitting project proposals outlining what they are hoping to achieve. We normally end up with about 50 or 60 submissions. We use a sophisticated review and assessment process to decide which projects we will take on, involving 10-15 people,” said Justin.
“The projects are quite short term, and each person might take on 3 projects every 6 months, doing one project for one or 2 days a week over that 6-month period.”
Demand for Research Software Expertise Is Increasing
Justin has seen the demand for research software capability increasing at CSIRO over the past few years.
The importance of research software is reflected in the many ways CSIRO deploys and nurtures this capability. CSIRO not only has research software capability in Scientific Computing Services – research software engineers and developers also sit inside research teams.
Dr Jens Klump, Group Leader for Exploration through Cover at CSIRO said, “The Discovery Program in CSIRO’s Mineral Resources Business Unit has invested heavily into growing its digital capability, both by employing research software engineers and by upskilling its scientific staff to become more familiar with digital tools.
“We have several research software engineers working on 2 ARDC-supported platform projects. They are providing significant input into the design of the Australian Scalable Drone Cloud, a platform for researchers using drones with scalable open source data processing, analysis, visualisation and publication tools. They are also supporting the ARDC project FAIMS 3.0, which is building a flexible, cross-platform app for producing custom electronic notebooks for data collection in the field.”
In addition to research software capability, Justin said, “There’s also a whole component of CSIRO that’s involved in upskilling researchers in this new digital realm.”
Dr Tom Honeyman, Program Leader for the ARDC’s Research Software Agenda, said, “CSIRO is an excellent example of what we are working on for all Australian research institutions – for the creation of research software to be recognised and nurtured as a vital capability for Australian research.”
Exciting Opportunities in Research Software
Justin’s group works on a wide range of projects. One that Justin particularly enjoyed working on recently involved albatross research.
“Researchers were trying to gauge the health of an albatross colony in the Bass Strait,” said Justin.
In person surveys of remote islands are costly, time consuming and challenging for both the researchers and the birds they are trying to observe, so researchers turned to taking high-resolution images of the Albatross rookery. The images provide a wealth of data but were time consuming to analyse manually. The researchers asked Scientific and Computing Services for help using machine learning to analyse the images and speed up the analysis.
“They were taking large gigapixel images of rookeries. In the images, everything was grey – birds, chicks, rocks, nests and often even the sky. They wanted to try using machine learning to count the number of nests in each image instead of using people power.
“We helped develop a prototype for the researchers involved in the survey, with the aim to reduce the manual work overall. After some tuning, the prototype’s bird detections were more accurate than nest detections and both progressively improved with more training.”
A Research Software Agenda for Australia
The report on research software capability in Australia is part of the ARDC’s broader agenda on research software in Australia, which was released in its final version in April 2022. One of the 3 goals of the agenda is to sustain research software. Creating and nurturing sustainable research software capability is key to this goal.
The ARDC is inviting managers or academic leads of teams with research software capability to participate in one of two national forums to discuss the questions arising from the report. Please register your interest in participating by filling out this form.
The ARDC is funded through the National Collaborative Research Infrastructure Strategy (NCRIS) to support national digital research infrastructure for Australian researchers.