Projects

As a Researcher Software Engineer I take part in various research software projects and deliver training courses. As much as I can, I'm getting involved with both the local academic community at Oxford and the wider, global research software engineering movement.

Current software projects

PyBaMM

Design and manufacturing of more efficient and resilient energy storage solutions is a key effort towards the development of sustainable energy systems. Beyond our laptops and phones, batteries allow the storage of energy from fluctuating renewable energy sources such as wind-farms. Increasingly, electricity-powered vehicles are becoming reality, from cars all the way to small planes!

PyBaMM is a python package for designing, simulating and analysing battery models. Development started in 2019 at the Mathematical Institute (University of Oxford) and the project was soon adopted as the common modelling framework for the Faraday Institution, the main entity supporting battery research in the UK. Battery models are traditionally solved using multi-scale modelling software such as COMSOL. In contrast, PyBaMM is a free (as in freedom) package available to all for use but also to contribute to. Beyond efficiently solving battery models, PyBaMM aims at becoming a community tool, developed by and for battery scientist.

stochare

Stochastic models (understand mathematical models of random processes) arise in many scientific fields: physics, chemistry, biology, finance... Stochrare aims at providing a toolbox to support research dealing with stochastic processes regardless of the application domain. For instance:

Although the package was initially developed with an out-of-equilibrium statistical physics point of view, it aims at providing tools to support research with stochastic processes in any of these fields.

Past software projects

Oxford Visual Perception Screen (OxVPS)

A large number of stroke patients - up to almost 80% - suffer visual impairment after stroke. Such impairment can include neglecting objects on one side of vision, visual hallucinations, difficulties recognising faces or objects and impaired motion perception. Only 20% of patients and carers, however, self-report these impairments and systematic screening is therefore essential.

Together with Fergus Cooper and Mihaela Duta, we teamed up with Dr Kathleen Van Cleef (Experimental Psychology, University Of Oxford) to turn a previously developed paper based assessment into a full-blown mobile app. And app can be used by clinicians in Acute Stroke Units with very limited time and provide automated assessment scoring and reporting of the results.

The app was developed in typescript using the ionic cross-platform framework, and primarily targets Android tablets. A number of specific development challenges had to be addressed, including:

We also implemented an automatic scoring system, with a results overview being automatically generated at the end of each assessment for the benefit of the clinician performing the screening. More complete data is securely stored in a cloud database, and can be downloaded for additional processing by Dr Vancleef and her team.

smif

Infrastructure systems (think water and energy supply, transport systems...) provide essential basic services to society but can also have harmful social and environmental impacts. Plans and decisions are subject to uncertain socio-economic, technological and environmental change and these complexities are best studied through simulation. smif[1,2] is a software framework for the integration of simulation models of infrastructure systems such as energy, transport, water and digital communications.

Smif enables the coupling and orchestration of an heterogeneous ensemble of models: each model has its own space and time resolution, set of inputs and outputs and data formats. To run under smif, models must be wrapped by a well-defined Python class, which provides a common interface. Smif’s purpose is to provide infrastructure systems modellers a tool capable of handling the execution of the models according to their dependencies, as well as the necessary data exchange between models, transforming data if necessary.

The development of smif is driven by the National Infrastructure Systems Model (NISMOD2), which consists of a collection of high-resolution infrastructure models of the United Kingdom. However, smif is a very general framework that can applied whenever models can be executed from Python, via import, binding, or call out to a command-line executable.

  1. Will Usher and Tom Russell | A Software Framework for the Integration of Infrastructure Simulation Models | Journal of Open Research Software | 2019 | (doi)

  2. Will Usher, Tom Russell, Roald Schoenmakers, Craig Robson, Fergus Cooper and Thibault Lestang | nismod/smif v1.2.0 | Zenodo | 2019 | (doi)

Community

The Oxford Code Review Network

As a researcher it's often difficult to feedback on your codes. Is your code readable and understandable by someone else? Are you missing out on some good programming practice(s)? Would somebody else have taken a different approach to solve the problem at hand?

These questions can be answered through engaging in regular code reviews, i.e. sitting down with a few colleagues and discussing a chunk of your code. It's a time for friendly conversation, debate and knowledge transfer that results and an improved code, strengthened team-spirit and better/wider software skills. Code reviews are fairly common in industry, but quasi non-existent in academia. In July 2020 I initiated the Oxford Code Review Network,to promote code reviewing in academia and facilitate code reviews between researchers across the University Of Oxford, by providing a (online) forum - for now a GitHub repo - for researchers across Oxford to get in touch and do code reviews together.

I'm convinced than making code reviewing a standard practice in academic research can have a tremendous impact on the quality of research software as a whole, not only resulting from the reviews themselves, but also by enabling the sharing of software skills and knowledge across levels of experience, backgrounds and research communities. I feel like this is much needed.

Reproducible research Oxford

I've been a Reproducible Research Oxford (RROx) fellow since May 2020. RROx is the local Oxford branch of the wider UK Reproducibility Network, and organisation of academics that works towards the culture change required to make open and reproducible research the norm.

RROx is a very active group, hosting/supporting various seminars, training events and grassroots initiatives such as reproducibiliTea Oxford, the Oxford Code Review Network and the Oxford Free and Open Source Software group.

Oxford Free/Open Source Software (OxFOSS)

Together with Laura Fortunato, Rowan Wilson and Malika Ihle, we initiated a local interest group on free (as in free speech) software. The motivation was simple: as supporters of free software, there was no local group to get involved with and meet other free software supporters/activists. So we decided to start our own.

OxFOSS officially started in October 2020 as a very informal biweekly meeting open to anybody who wants to talk free software. Meeting are a combination of relaxed chats on a predefined topic and invited talks. We hope that these will foster a strong and friendly free software community in Oxford, that can give free software more visibility for academics but also outside the University.

If you're interested, you can subscribe to the OxFOSS mailing list and/or join our Matrix room. If you're not working at the University Of Oxford, feel free to get in touch with me and I can add you to the room.

Junior Research Fellowship at Kellogg College

In may 2020 I was awarded a Junior Research Fellowship from Kellogg College, starting October 2020. Despite the pandemic, I'm very excited to take part in the College's life and connect with students and other fellows. I see this fellowship as a great opportunity to make research software engineering more visible but also learn more about where and how is research software developed across research fields and communities. I believe than research software engineers can bring a lot to colleges, and I am very grateful to Kellogg College for giving me the opportunity to demonstrate this!

Courses

A crucial part of our mission at Oxford Research Software Engineering is to deliver training courses to the local research community. You can find the list of courses that we run regularly here. Feel free to get in touch if you'd like us to run a course for your team/lab/department. Among them, I'm responsible for the following:

Python packaging

A three hours course on organising and sharing python projects, using python packages.

Participants start with a couple of messy analysis scripts that they incrementally turn into a full blown python package they finally release on PyPI (well, TestPyPI to be accurate).

This course gives researchers who write python to tools to organise their projects in a way that makes reusing code straightforward, but also efficient and sustainable. It also demonstrates that sharing this code with others - beyond exchanging USB sticks - is very simple. Anyone can make their python project "pip installable".