Scalable model agnostic GPU framework for Bayesian Deep Learning and Bayesian optimization (EPSRC CDT in Distributed Algorithms)


This studentship has been developed by the University of Liverpool and STFC’s Hartree centre in partnership with Rolls Royce.

This project will develop Bayesian Deep Learning algorithms using sparse and large datasets. The development aims to take advantage of parallel architectures in a cloud-based environment using open-source tools.

Deep learning    is receiving a great amount of attention due to its ability to perform classification tasks that can exceed human-level performance. However, modern deep learning algorithms are usually not able to properly quantify uncertainty. Without this key ingredient, there’s a serious risk of blindly assuming the correctness of the underlying models. This can have costly consequences when solving realistic, industrially relevant problems. Bayesian deep learning can model complex tasks by assigning probability distributions to the corresponding model parameters. This way, an estimate of the accuracy of the output is provided, and thus a level of confidence in the prediction allows to quantify the quality of the proposed solutions.

The key challenges in this work are the intensive computational requirements imposed by the implementations of deep learning. One possible solution is the adoption of general-purpose Graphical Processing Units (GPUs). However, it is not trivial to determine the correct GPU configuration to achieve an optimal implementation. Additionally, the current algorithms might not scale to GPU clusters. Therefore, this project requires the development of efficient and scalable algorithms able to exploit the computational power of modern hardware. It is expected that the research in this project will have applications to computer vision. Proof of concepts will be produced, firstly to demonstrate the GPU migration code and secondly to demonstrate the capabilities of the Bayesian Deep Learning algorithms that will be developed.

This project is part of the EPSRC Funded CDT in Distributed Algorithms: The What, How and where of Next-Generation Data Science.

The University of Liverpool is working in partnership with the STFC Hartree Centre and other industrial partners from the manufacturing, defence and security sectors to provide a 4 year innovative PhD training course that will equip over 60 students with the essential skills needed to become future leaders in data science, be it in academia or industry.

Every project within the centre is offered in collaboration with an Industrial partner who as well as providing co-supervision will also offer the unique opportunity for students to access state of the art computing platforms, work on real world problems, benchmarking and data. Our graduates will gain unparalleled experiences working across academic disciplines in highly sought-after topic areas, answering industry need.

As well as learning from academic and industrial world leaders, the centre has a dedicated programme of interdisciplinary research training including the opportunity to undertake modules at the global pinnacle of Data science teaching.  A large number of events and training sessions are undertaken as a cohort of PhD students, allowing you to build personal and professional relationships that we hope will lead to research collaboration either now or in your future.

The learning nurtured at this centre will be based upon anticipation of the hardware recourses arriving on desks of students after they graduate, rather than the hardware available today.

For informal enquires please contact Dr Edoardo Patelli or

To apply for this Studentship please submit an application for an Engineering PhD via our online platform ( and provide the studentship title and supervisor details when prompted. Should you wish to apply for more than one project, please provide a ranked list of those you are interested in.

For a full list of the entry criteria and a recruitment timeline (including interview dates etc), Please see our website



Jony Castagna - STFC Hartree Centre

I joined the Science and Technology Facilities Council (STFC) Hartree Centre in 2016 and since then I have enjoyed my work more and more. I am a computational scientist with a strong passion for high performance computing (HPC), usually oriented to the Simulation of Turbulent Flows using Computational Fluid Dynamics… But let me just say it: I love programming GPUs! CUDA is my favourite language, but I recently start to use OpenACC more due its portability to other platforms. 

I am currently working on several projects. As a software developer for the European HPC Centre of Excellence (E-CAM), I mainly help scientists to achieve high performance with their code as well as reorganising their software using best practice guidelines. I ported the DL_MESO code for mesoscale simulation on NVIDIA GPU, achieving a nice scaling on the latest PRACE supercomputer, Piz Daint in Switzerland.  

Going towards smaller scales, I work in collaboration with the IBM Research team at the Hartree Centre. We are trying to port a quantum molecular dynamic code named QDRUDE, which simulates life science systems, to GPUs. We mainly run our simulations on the Hartree Centre supercomputer Panther: an IBM + NVIDIA architecture.

Finally, working with GPUs since 2010, how could I not end up in Deep Learning? This new fascinating world has recently captured me… and transformed me into an NVIDIA Ambassador here at STFC! So, while I enjoy running NVIDIA Deep Learning Institute courses ranging from CUDA to Deep Learning for Computer Vision, my main focus stays on applied HPC for scientific research, mainly using future computing systems like hybrid CPU-GPU architectures where integration between artificial intelligence (AI) and traditional HPC science is merged together.