My proposed Fellowship will revolutionise the use of High Performance Computing (HPC) within The University of Sheffield by changing perceptions of how people utilise software and are trained and supported in writing code which scales to increasingly large computer systems. I will provide leadership by demonstrating the effectiveness of specific research software engineer roles, and by growing a team of research software engineer at The University of Sheffield in order to accommodate our expanding programme of research computing. I will achieve this by: 1) developing the FLAME and FLAME GPU software to facilitate and demonstrate the impact of Graphics Processing Unit (GPU) computing on the areas of complex systems simulation; 2) vastly extending the remit of GPUComputing@Sheffield to provide advanced training and research consultancy, and to embed specific software engineering skills for high-performance data parallel computing (with GPUs and Xeon Phis) across EPSRC-remit research areas at The University of Sheffield. My first activity will enable long-term support of the extensive use of FLAME and FLAME GPU for EPSRC, industry and EU-funded research projects. The computational science and engineering projects supported will include those as diverse as computational economics, bioinformatics and transport simulation. Additionally, my software will provide a platform for more fundamental computer science research into complexity science, graphics and visualisation, programming languages and compilers, and software engineering. My second activity will champion GPU computing within The University of Sheffield (and beyond to its collaborators and industrial partners). It will demonstrate how a specific area of research software engineering can be embedded into The University of Sheffield, and act as a model for further improvement in areas such as research software and data storage. I will change the way people develop and use research software by providing training to students and researchers who can then embed GPU software engineering skills across research domains. I will also aid researchers who work on computationally demanding research by providing software engineering consultancy in areas that can benefit from GPU acceleration, such as, mobile GPU computing for robotics, deep neural network simulation for machine learning (including speech, hearing and Natural language processing) and real time signal processing. The impact of my Fellowship will vastly expand the scale and quality of research computing at The University of Sheffield, embed skills within students and researchers (with long-term and wide-reaching results) and ensure energy-efficient use of HPC. This will promote the understanding and wider use of GPU computing within research, as well as transitioning researchers to larger regional and national HPC facilities. Ultimately my research software engineer fellowship will facilitate the delivery of excellent science whilst promoting the unique and important role of the Research Software Engineer.

Humans interact with tens of objects daily, at home (e.g. cooking/cleaning) or outdoors (e.g. ticket machines/shopping bags), during working (e.g. assembly/machinery) or leisure hours (e.g. playing/sports), individually or collaboratively. When observing people interacting with objects, our vision assisted by the sense of hearing is the main tool to perceive these interactions. Let's take the example of boiling water from a kettle. We observe the actor press a button, wait and hear the water boil and the kettle's light go off before water is used for, say, preparing tea. The perception process is formed from understanding intentional interactions (called ideomotor actions) as well as reactive actions to dynamic stimuli in the environment (referred to as sensormotor actions). As observers, we understand and can ultimately replicate such interactions using our sensory input, along with our underlying complex cognitive processes of event perception. Evidence in behavioural sciences demonstrates that these human cognitive processes are highly modularised, and these modules collaborate to achieve our outstanding human-level perception. However, current approaches in artificial intelligence are lacking in their modularity and accordingly their capabilities. To achieve human-level perception of object interactions, including online perception when the interaction results in mistakes (e.g. water is spilled) or risks (e.g. boiling water is spilled), this fellowship focuses on informing computer vision and machine learning models, including deep learning architectures, from well-studied cognitive behavioural frameworks. Deep learning architectures have achieved superior performance, compared to their hand-crafted predecessors, on video-level classification, however their performance on fine-grained understanding within the video remains modest. Current models are easily fooled by similar motions or incomplete actions, as shown by recent research. This fellowship focuses on empowering these models through modularisation, a principle proven since the 50s in Fodor's Modularity of the Mind, and frequently studied by cognitive psychologists in controlled lab environments. Modularity of high-level perception, along with the power of deep learning architectures, will bring a new understanding to videos analysis previously unexplored. The targeted perception, of daily and rare object interactions, will lay the foundations for applications including assistive technologies using wearable computing, and robot imitation learning. We will work closely with three industrial partners to pave potential knowledge transfer paths to applications. Additionally, the fellowship will actively engage international researchers through workshops, benchmarks and public challenges on large datasets, to encourage other researchers to address problems related to fine-grained perception in video understanding.

Bayesian inference is a process which allows us to extract information from data. The process uses prior knowledge articulated as statistical models for the data. We are focused on developing a transformational solution to Data Science problems that can be posed as such Bayesian inference tasks. An existing family of algorithms, called Markov chain Monte Carlo (MCMC) algorithms, offer a family of solutions that offer impressive accuracy but demand significant computational load. For a significant subset of the users of Data Science that we interact with, while the accuracy offered by MCMC is recognised as potentially transformational, the computational load is just too great for MCMC to be a practical alternative to existing approaches. These users include academics working in science (e.g., Physics, Chemistry, Biology and the social sciences) as well as government and industry (e.g., in the pharmaceutical, defence and manufacturing sectors). The problem is then how to make the accuracy offered by MCMC accessible at a fraction of the computational cost. The solution we propose is based on replacing MCMC with a more recently developed family of algorithms, Sequential Monte Carlo (SMC) samplers. While MCMC, at its heart, manipulates a single sampling process, SMC samplers are an inherently population-based algorithm that manipulates a population of samples. This makes SMC samplers well suited to the task of being implemented in a way that exploits parallel computational resources. It is therefore possible to use emerging hardware (e.g., Graphics Processor Units (GPUs), Field Programmable Gate Arrays (FPGAs) and Intel's Xeon Phis as well as High Performance Computing (HPC) clusters) to make SMC samplers run faster. Indeed, our recent work (which has had to remove some algorithmic bottlenecks before making the progress we have achieved) has shown that SMC samplers can offer accuracy similar to MCMC but with implementations that are better suited to such emerging hardware. The benefits of using an SMC sampler in place of MCMC go beyond those made possible by simply posing a (tough) parallel computing challenge. The parameters of an MCMC algorithm necessarily differ from those related to a SMC sampler. These differences offer opportunities for SMC samplers to be developed in directions that are not possible with MCMC. For example, SMC samplers, in contrast to MCMC algorithms, can be configured to exploit a memory of their historic behaviour and can be designed to smoothly transition between problems. It seems likely that by exploiting such opportunities, we will generate SMC samplers that can outperform MCMC even more than is possible by using parallelised implementations alone. Our interactions with users, our experience of parallelising SMC samplers and the preliminary results we have obtained when comparing SMC samplers and MCMC make us excited about the potential that SMC samplers offer as a "New Approach for Data Science". Our current work has only begun to explore the potential offered by SMC samplers. We perceive significant benefit could result from a larger programme of work that helps us understand the extent to which users will benefit from replacing MCMC with SMC samplers. We propose a programme of work that combines a focus on users' problems with a systematic investigation into the opportunities offered by SMC samplers. Our strategy for achieving impact comprises multiple tactics. Specifically, we will: use identified users to act as "evangelists" in each of their domains; work with our hardware-oriented partners to produce high-performance reference implementations; engage with the developer team for Stan (the most widely-used generic MCMC implementation); work with the Industrial Mathematics Knowledge Transfer Network and the Alan Turing Institute to engage with both users and other algorithmic developers.

This proposal brings together communities from the UK Turbulence Consortium (UKTC) and the UK Consortium on Reacting Flows (UKCRF) to ensure a smooth transition to exascale computing, with the aim to develop transformative techniques for future-proofing their production simulation software ecosystems dedicated to the study of turbulent flows. Understanding, predicting and controlling turbulent flows is of central importance and a limiting factor to a vast range of industries. Many of the environmental and energy-related issues we face today cannot possibly be tackled without a better understanding of turbulence. The UK is preparing for the exascale era through the ExCALIBUR programme to develop exascale-ready algorithms and software. Based on the findings from the Design and Development Working Group (DDWG) on turbulence at the exascale, this project is bringing together communities representing two of the seven UK HEC Consortia, the UKTC and the UKCTRF, to re-engineer or extend the capabilities of four of their production and research flow solvers for exascale computing: XCOMPACT3D, OPENSBLI, UDALES and SENGA+. These open-source, well-established, community flow solvers are based on finite-difference methods on structured meshes and will be developed to meet the challenges associated with exascale computing while taking advantage of the significant opportunities afforded by exascale systems. A key aim of this project is to leverage the well-established Domain Specific Language (DLS) framework OPS and the 2DECOMP&FFT library to allow XCOMPACT3D, OPENSBLI, UDALES and SENGA+ to run on large-scale heterogeneous computers. OPS was developed in the UK in the last ten years and it targets applications on multi-block structured meshes. It can currently generate code using CUDA, OPENACC/OPENMP5.0, OPENCL, SYCL/ONEAPI, HIP and their combinations with MPI. The OPS DSLs' capabilities will be extended in this project, specifically its code-generation tool-chain for robust, fail-safe parallel code generation. A related strand of work will use the 2DECOMP&FFT a Fortran-based library based on a 2D domain decomposition for spatially implicit numerical algorithms on monobloc structured meshes. The library includes a highly scalable and efficient interface to perform Fast Fourier Transforms (FFTs) and relies on MPI providing a user-friendly programming interface that hides communication details from application developers. 2DECOMP&FFT will be completely redesigned for a use on heterogeneous supercomputers (CPUs and GPUS from different vendors) using a hybrid strategy. The project will also combine exascale-ready coupling interfaces, UQ capabilities, I/O & visualisation tools to our flow solvers, as well as machine learning based algorithms, to address some of the key challenges and opportunities identified by the DDWG on turbulence at the exascale. This will be done in collaboration with several of the recently funded ExCALIBUR cross-cutting projects. The project will focus on four high-priority use cases (one for each solver), defined as high quality, high impact research made possible by a step-change in simulation performance. The use cases will focus on wind energy, green aviation, air quality and net-zero combustion. Exascale computing will be a game changer in these areas and will contribute to make the UK a greener nation (The UK commits to net zero carbon emissions by 2050). The use cases will be used to demonstrate the potential of the re-designed flow solvers based on OPS and 2DECOMP&FFT, for a wide range of hardware and parallel paradigms.

Over the past decades, computer simulations have played an increasingly important role in design of numerous complex systems. In particular, computer simulations have played a pivotal role in aerodynamic and structural design of aircraft. It is becoming apparent, however, that current generation software packages used for aerodynamics design are not fit for purpose. Newer software is required, that can make effective use of current and future computing platforms, to perform highly accurate so called 'scale-resolving' simulations of air flow over complex aircraft configurations. Such capability would lead to design of more efficient and capable aerospace technology. In particular, it would greatly improve design of next generation Unmanned Aerial Vehicles (UAVs), which in the coming decades are set to have a significant impact on our society, playing key roles in areas such as defense, border security, search and rescue, farming, fishing, cargo transport, wireless communications, and weather monitoring. The primary objectives of this research are to i.) develop software that can effectively leverage capabilities of current and future computing platforms (with many thousands or even millions of computing cores) to undertaken hitherto intractable simulations of airflow over complex UAV configurations ii.) test and demonstrate cutting edge functionality of this software, iii.) translate the technology to industry, such that it can be used to facilitate design of next generation UAVs. The research program will be lead by Dr. Peter Vincent (a Lecturer in the department of Aeronautics at Imperial College London). It will be undertaken in collaboration with various industrial partners including BAE Systems, NASA Glenn, Nvidia, and Zenotech, and with various academic partners including Stanford University, UC Berkeley, University of Swansea, and University of Utah. This assembled team of project partners, comprising a selection of the world's leading companies, and elite research institutions, will ensure the project successfully delivers its objectives.