HPC/AI Specialist

Overview of Responsibilities
Who are we looking for?

We are seeking applicants with experience in delivering novel HPC and AI solutions at scale (for science or engineering) and a keen interest in delivering solutions that will leverage both at extreme scale, i.e., exploiting the so-called “convergence” or AI and HPC at the exascale. We expect applicants to have a background in science, computer-science, or engineering, having worked upon large-scale, remote systems for large scale simulation and with Big Data.

Successful candidates must be able to demonstrate that they have developed software / solutions (e.g., algorithms) that exploit large, distributed HPC systems (i.e., relying upon technologies such as MPI, OpenMP etc. as well as accelerator solution (i.e., GPUs). Experience in using some of the more novel architectures is highly desirable (e.g., Graphcore, Cerebras, SambaNova, A64FX, ARM etc.).

Experience can be via academic research / training (e.g., MSc. or PhD. research in HPC/AI) or from working in academic scientific research or in industry. Successful applicants will have a passion for, and demonstrable experience of having helped take solutions through from concept to production and a keen interest in working with stakeholders from requirements capture through to service delivery.

The role will carry with it a significant level of autonomy, allowing the successful candidate to be creative and grow this area. Applicants with an extensive network and/or interest in growing collaborations are particularly encouraged.

The role involves spanning one (or more) of the following areas:

• Exascale targeted algorithms – developing new numerical methods to achieve performance and scalability, esp. in deploying AI methods to model sub-grid physics or e.g., for providing coarse solvers for the parallel in time methods or for pre-conditioning.
• Novel hardware exploitation – working with hardware OEMs to exploit advances in exascale targeted architectures, for simulating complex systems or for real world systems control problems (e.g., exploiting architectures such as Cerebras, Graphcore, SambaNova / data-flow programming etc.).
• Development of surrogate modelling / emulation technology – Gaussian Process Emulation, Polynomial Chaos, AI based surrogates (PINNs, FNO, PINO) etc.
• Uncertainty Quantification (UQ) – Making our simulation code base “actionable” by wrapping our codes (or embedding within them) UQ infrastructure, exploiting advances in UQ such as imprecise probability.
• Development of design of experiments for simulation / sequential design tools/methods – ensuring that the execution of simulation at scale is optimised to guarantee that maximal new information is extracted per £ invested in HPC infrastructure – computational steering, instrumenting simulations with telemetry, in-situ visualisation methods etc.
• Ensuring that our codes are performant and scalable and building performance portability into our codes and workflows – performing roofline analysis, porting codes to new architectures, applying performance portability methods/tools such as Kokkos and SYCL/oneAPI to our codebase to ensure we can exploit supercomputers across the fusion / UKRI landscape.

Successful applicants will be expected to have demonstrable advanced skills in software engineering, esp. around scripting and languages such as C/C++ and/or FORTRAN and ideally exposure to various of the ubiquitous AI platforms (TensorFlow, PyTorch etc.).

A key requirement for the role is that the applicant possess the ability to work effectively within a rapidly growing, diverse team. The successful candidate will have a passion for nurturing science and technological excellence. If you have a background in HPC/AI, a creative problem-solving mind, knowledge of modern software engineering processes and best practice and are excited about helping to solve one of the world’s most pressing grand challenges (to deliver UK Net Zero by 2050), then we are keen to hear from you!

What will you be responsible for?

Key areas of responsibility will include the following:

• Developing advanced simulation tools with a focus upon extreme scale simulation and exploitation of AI/HPC convergence, from back-end to front-end, through a process of co-design and regular engagement with stakeholders/users.
• Collaborating with pioneers developing novel HPC architectures (e.g., Graphcore, Cerebras etc.) to co-design fusion simulation capability for the exascale.
• Providing user support and assisting in the production of training materials so that simulation codes / tools are suitable for long-term use and receive the widest possible adoption across UKAEA and its growing academic and industrial partnerships.
• Acting as an “authority” or point of contact for all matters surrounding the exploitation of advanced computing to further fusion science and engineering.
• Remaining current in scientific and technical software engineering and around advances in the HPC / AI space, promoting best practice both internally and among external partners.
• Representing the UK at International Conferences/workshops, presenting one’s own work or that of team members.
• Writing and publishing high quality research in internationally recognised peer reviewed journals or popular articles for industry journals such as HPCWire.
• Helping to promote a healthy working environment and team-focused work ethic, centred upon continuous improvement and a core ethos of inclusivity and diversity.

Please note: We may consider the post being based at alternative locations to CCFE, provided those locations are part of UKAEA’s growing network of collaborations (e.g., Daresbury Laboratory).

Please note that this role is split over two levels and the Level 4 offers a salary of £36,397 – £38,620 (inclusive of MPP) and the Level 5 offers a salary of £49,396 – £52,293 (inclusive of MPP). Your level will be assessed as part of the interview process.

Qualifications
Essential skills, experience and competence required

• At least a Master’s degree in a computational, mathematical or physical science discipline or equivalent experience (including a knowledge of the various methods for solving systems of coupled Partial Differential Equations).
• Background/interest in the power of advanced computing for delivering actionable, i.e., predictive capability in the modelling of complex, strongly coupled non-linear systems.
• Programming skills in C/C++ and or FORTRAN, python (or equivalent, e.g., Julia) and associated scripting languages.
• Knowledge of modern software engineering processes and best practice, e.g., testing frameworks, Git etc.
• Familiarity with the libraries, tools, methods required for HPC (e.g., distributed memory computing (MPI), shared memory systems, OpenMP, accelerator programming, e.g., CUDA, OpenACC etc.).
• At least a basic knowledge of AI platforms, e.g., TensorFlow, PyTorch etc.
• Good experience of the UNIX operating system, particularly the command line interface.

Desirable skills, experience, and competence

• A background in plasma physics, fluid dynamics or other field requiring the simulation of complex systems.
• Knowledge/experience of one or more of the currently available performance portability frameworks, e.g., Kokkos, OpenMP 5, SYCL/oneAPI, RAJA etc.
• Knowledge of one or more of the available UQ platforms for UQ at extreme scale, e.g., VECMA, ExaQUte etc.
• Experience of the high order spectral-hp method or implicit Particle In Cell methods.
• Experience of having coupled particle methods to fluid solvers.

Additional Information
What we offer

Our comprehensive benefits package includes a vibrant culture committed to being fully inclusive, an Employee Assistance Programme and trained Mental Health First Aiders, Flexible working options including family friendly policies, Emergency paid leave, Generous annual leave allowance (30.5 days) increased with length of service, wide range of career development opportunities, outstanding defined benefit pension scheme; Corporate bonus scheme and a Relocation allowance (if eligible).

We welcome applications from under-represented groups, particularly from individuals from black and other ethnic minority backgrounds, including nationality and citizenship, people with disabilities and women. The Executive team with support from our ‘Equality, Diversity and Inclusion’ (EDI) Partner and Inclusion Ambassadors, is actively promoting EDI and taking steps to increase the diversity of our people through reinforcing best practice in recruitment and selection, and revaluating approaches where it is clear we can remove barriers to success.

Please be advised that this vacancy is due to close on 08/08/2022 . We may close earlier than this date if large or sufficient numbers of applications are received.

Please note all employees working at the UK Atomic Energy Authority will be required to complete an online Disclosure Certificate application as part of their clearance – The Disclosure & Barring Service (DBS) checks will show the details of all current criminal convictions or will confirm that there are no such convictions.

Note for applicants who are applying for employment with UKAEA either from outside the United Kingdom or who have spent time outside the United Kingdom in the last five years. Please visit  https://www.gov.uk/government/publications/criminal-records-checks-for-overseas-applicants . If the country you are resident in or have resided in is not listed on this website or that the UK Government is not aware of how to obtain a criminal records check from that state then we are unable to process your application.