Recruitment

Prospective Graduate Students

Thank you for your interest! We are looking for Ph.D. candidates in areas of Cyber-Physical Systems, Mobile/Wearable Systems, Edge-AI and On-chip Intelligence.
Students with either a CS or EE background are highly encouraged to apply.

Candidates are expected to have:

(1) a good background in either computer hardware, mobile systems, or AI,

(2) a strong motivation towards academic excellence,

(3) and should be prepared for 3-4 years of concentrated work.

The following backgrounds are not essential but would be a big plus:

(4) Graduated with a Master's Degree from accredited colleges and universities.

(5) Published research articles as first author.

(6) Experience in developing cyber-physical, mobile, or wearable systems. 

(7) Experience in digital signal processing and/or machine learning.

Students in our team will build a diverse skill set in both system implementation and theoretical modeling. They will also acquire a broad knowledge of physiological sensing, mobile healthcare, wearable computing, edge AI, etc. Because of such unique strength, our team has publications and awards in highly selective venues such as MobiCom, MobiSys, SenSys, UbiComp, ACM/IEEE Transactions, etc. We also have collaborations with people in Google, Oxford, Cambridge, Imperial, Uni. of Massachusetts, and Uni. of Colorado that could help open up your research career after graduation.

Please refer to our publication list to get a general understanding of our research topics.
Please send your pre-application package, including a CV, a sample of publications, and IELTS/TOEFL, to Dr. Nhat (Nick) Pham (phamn@cardiff.ac.uk) if you are interested.

PhD Project Topics

(1) (Deadline: 9/1/2026) High-Resolution On-Device Machine Vision using a Microcontroller and a Neural Processing Unit for Scalable, Low-Power Biodiversity Sensing [FindaPhD]

Camera systems for automated monitoring of insects are field-deployed devices that autonomously capture pictures of insects (e.g., moths, butterflies...), which are then processed by multi-stage machine-learning pipelines that detect and classify individual insects. This is a challenging task due to the small size of many insects, the enormous number of different species – many of which look similar – and a strong class imbalance in available training data. Therefore, typically high-resolution cameras are used in conjunction with high-performance (GPU) cloud computing. This is costly and poses the challenge of transferring large data packages from device to cloud. 

To address this problem, this PhD will design and validate a microcontroller-based (MCU-based) device for automated monitoring of insects that captures high-resolution images, detects individuals, and provides hierarchical identification on device while running from small batteries. We target the latest MCU platforms with neural processing units (NPU), and co-design hardware, capture scheduling and embedded vision pipelines to maximise species-level recall per watt. Recent MCU platforms with NPUs make it feasible to run vision models on high-resolution inputs, but only with careful pipeline and model co-design. The PhD tackles three intertwined challenges: (1) an embedded pipeline that capture pictures, performs low-power region-of-interest detection, and then hierarchical classification on crops; (2) resource-aware machine learning compiled for MCU NPUs; (3) optimisation of energy, throughput, and accuracy. Outputs will include: (i) an open hardware/firmware reference design; (ii) a compiler-friendly model zoo with reproducible energy/latency/accuracy benchmarks; (iii) a deployment toolkit for real-world trials. 

The overarching question is how close can MCU+NPU hardware get to GPU baselines for small-object detection and fine-grained classification under tight energy budgets? A team of supervisors with relevant experience will provide training on aspects such as on-device machine learning and wildlife monitoring hardware and a lab with equipment for prototyping and compute. 

The project advances high-resolution on-device vision on MCUs for ecology, enabling scalable monitoring across many sites. Insect monitoring is of significant interest to many stakeholders since insects underpin ecosystems as pollinators, decomposers, and prey; their abundance and activity are widely used as indicators of biodiversity. 

This project asks whether high-resolution visual recognition can be done credibly on microcontrollers. The scientific contribution goes beyond engineering a device to actual step-changes: For example, a new theory for resource-constrained vision, formalising species-level recall per watt and developing laws that link image resolution, model capacity, and NPU memory/compute. This yields principled design rules for on-device biodiversity sensing.

Supervision team.

Dr Jonas Beuchert, Dr Nhat (Nick) Pham, and Dr Tom August (Centre for Ecology & Hydrology)

(2) Enabling microwatt-level brain-inspired neuromorphic AI with compound semiconductors [Available for Compound Semiconductor CDT Students]

The wearable and implantable devices experienced significant growth in recent years, reaching over $186 billion globally by 2030. Soon, there could be hundreds of intelligent devices that weave themselves into the fabrics of everyday wearables, providing us with various information about our health and well-being. To process this tremendous amount of information, state-of-the-art embedded AI systems are struggling to keep up with a limited amount of battery power.

The human brain is the most efficient computer, with 1015 synapses and 1011 neurons, while consuming less than 20W. However, creating an efficient on-chip neuromorphic AI system is not a trivial task due to the inherent limitations of direct memory access, fan-ins and fan-outs in conventional CMOS transistors. In this project, the student will work with experts in semiconductors and AI at Cardiff, UCL, and Google to:

(1) Enable efficient hardware that can provide direct memory access and highly connected neurons and synapses, based on compound semiconductors (e.g., InP, or GaAs).

(2) Devise a novel framework to run advanced neuromorphic AI algorithms, such as reservoir computing or spike-timing-dependent-plasticity (STDP).

(3) Develop a prototype and evaluate the developed system in clinical and industrial settings to study the usability of the proposed solution. 

References.

[1] Marković, D., Mizrahi, A., Querlioz, D., & Grollier, J. (2020). Physics for neuromorphic computing. Nature Reviews Physics, 2(9), 499-510.

[2] Schuman, C. D., Kulkarni, S. R., Parsa, M., Mitchell, J. P., Date, P., & Kay, B. (2022). Opportunities for neuromorphic computing algorithms and applications. Nature Computational Science, 2(1), 10-19.

[3] Kim, Seung Ju, et al. "2D materials-based 3D integration for neuromorphic hardware." npj 2D Materials and Applications 8.1 (2024): 70.

[4] Boybat, I., Le Gallo, M., Nandakumar, S. R., Moraitis, T., Parnell, T., Tuma, T., ... & Eleftheriou, E. (2018). Neuromorphic computing with multi-memristive synapses. Nature communications, 9(1), 2514.

Supervision team.

Dr Nhat (Nick) Pham, (Lead supervisor), Dr Chris Lu (University College London), Dr Seobin Jung (Google Research)

Funding

There are various sources of funding for PhD projects from the School of Computer Science and Informatics and UK research councils. Most of the studentships will cover the tuition fee and living expenses for 3.5 years (no tuition fee after 3 years). Prospective candidates are always welcome if they can secure their own funding sources.

Application deadlines

There are four intakes for the Ph.D. program each year, i.e., October, January, April, and July. However, if you want to apply for School scholarships, there is only one intake in October with the application deadline is 13 March. Details are available on the School's website.

UK research councils' studentships

EPSRC Doctoral Training Partnership to develop Distributedly Self-powered Wearables (Project 2):  ~£90k, 2024 - 2028, 100% home-fee + £18.6k PhD annual tax-free stipend for 3.5 years. Additional funding to supplement international fees is also possible.

School Scholarships (UK + International)

The School offers fully-funded scholarships (Deadline: 13 March) for outstanding applicants (both UK and international students). The scholarships are competitive, so it will always be good to reach out to me early to prepare the application.

China Scholarship Council (CSC) + Cardiff University 

Deadline: 30 November 2023 with the School of Computer Science and Informatics @ Cardiff (Apply here). The China Scholarship Council (CSC) predominantly funds Chinese university students to study overseas for PhD degrees in the fields of science and technology, but will also consider disciplines in the humanities. CSC awards typically cover airfares and living costs for up to four years. The School will nominate up to 10 candidates.

Commonwealth Scholarship

Commonwealth Scholarships enable talented and motivated individuals to gain the knowledge and skills required for sustainable development and are offered to citizens from low and middle-income Commonwealth countries. More info and application can be found here.

Teaching Associate + Part-time PhD (UK students or visa holders)

The School also often recruits teaching associate posts with a part-time PhD opportunity (6-year program). As it is restricted by Visa requirements, it is currently open to candidates who already have work permits in the UK. The TA posts will usually be advertised in August.

Current TA advertisement (closed on 22/9/2023)