Michael Ajao-Olarinoye

Applied machine learning · Generative AI · Scientific modelling

Michael Ajao-Olarinoye

Research Fellow, Coventry University
Lecturer in Computing, QAHE

PhD · AFHEA

Google ScholarORCIDGitHubLinkedInCoventry Pure

Download CVPublications

Portrait of Dr Michael Ajao-Olarinoye

About

I am a Research Fellow at Coventry University and a Lecturer in Computing with QAHE. My work covers applied machine learning, generative AI and scientific modelling.

I lead AI research on EVOLVE AI, which models student engagement and academic risk to support early intervention. Before that I led machine learning and generative AI research on HOPE-MOVE, an Innovate UK digital health project, building participant dropout prediction and language models for facilitator support.

I also collaborate on physics-informed neural networks for cardiovascular blood flow, work that began during my PhD in Computational Science and Mathematical Modelling at Coventry.

News

  1. 20 Oct 2026

    Invited talk at the launch of the Centre for Doctoral Training in Artificial Intelligence, Coventry University: “From Student to Success: An AI Alumni Journey”.

  2. Sep 2026

    Endorsed by the Royal Academy of Engineering under UK Global Talent, Exceptional Promise.

  3. Sep 2026

    Started as Research Fellow on EVOLVE AI at Coventry University.

  4. Sep 2026

    New paper in Physics of Fluids on wall shear stress in coronary arteries and grafts.

  5. 10 Sep 2026

    Presented first-author work on NHS critical-care surge allocation at UKCI 2026, Coventry.

  6. Aug 2026

    New SSRN preprint on weather-aware home energy management with language models.

  7. Jul 2026

    Two new arXiv preprints, on language-model agents for home energy and on retrieval for document question answering.

  8. Mar 2026

    Paper on double descending thoracic aortic aneurysms accepted by Pramana – Journal of Physics.

Research

Physics-informed models of blood flow

Neural networks that learn the physics of blood flow from simulation data, so wall shear stress in arteries, aneurysms and bypass grafts can be estimated in seconds rather than hours.

With collaborators at NUST, Pakistan, and colleagues in CSMM, Coventry.

See the papers (5)
CFD and PINN wall shear stress maps side by side for one Marfan aneurysm case, in two views, at systole and diastole.
Physics of Fluids, 2025, Fig. 7(a, b) · CC BY 4.0CFD and PINN wall shear stress side by side for one Marfan aneurysm case, at systole and diastole.

Scientific machine learning

Comparing classical numerical schemes with physics-informed networks on dam-break and shallow-water flows, and joining compartment models to neural networks.

With colleagues at NUST Islamabad and Coventry University.

See the papers (4)
Flow diagram of the compartmental epidemic model, from susceptible through exposed, infectious, hospital and critical care to recovered or deceased.
Hybrid PINN–SEIRD modelThe compartmental model behind the hybrid physics-informed network.

Applied and generative AI

Language-model agents and retrieval systems, and machine learning for student engagement and digital health.

Coventry University and research co-authors.

See the papers (2)
System diagram: the Hope Move platform feeds a facilitator dashboard, which calls a risk engine and a drafting engine. A facilitator reviews, edits and sends each reply.
HOPE-MOVERisk and drafting engines behind a facilitator dashboard. A person reviews and sends every reply.
Illustration of a person with a knee brace working at a laptop at home.
HOPE-MOVE · IllustrationDigital support for people managing long-term conditions at home.

Graph and hybrid models for health systems

Graph attention networks for regional epidemics, and physics-informed models joined to an optimiser that allocates NHS critical-care capacity.

Developed during my PhD at Coventry University.

See the papers (4)
Pipeline diagram running from regional NHS England data through a physics-informed encoder and a temporal model to an exact regional allocation.
UKCI 2026, in pressFrom regional NHS data, through a physics-informed encoder, to an exact regional allocation.
Flow diagram of MSAGAT-Net: input and adjacency matrices pass through feature extraction, EAGAM, DMTFM with four dilation rates and PPRM to the output, with a skip connection from input to PPRM.
MSAGAT-Net, Fig. 1Input and adjacency pass through feature extraction, EAGAM, DMTFM and PPRM, with a skip connection from input to PPRM.

Publications

Newest first, grouped by status. Filter by topic or status, and copy a citation from any entry.

Published

  1. Three views of a patient-specific coronary model coloured by wall shear stress: CFD, PINN and the absolute error between them.

    PublishedNew2026

    Wall shear stress in coronary arteries and grafts using computational fluid dynamics and physics-informed neural networks

    Ur Rehman, M. A., Ekici, Ö., Erdener, Ş. E., Ajao-Olarinoye, M., Kuchumov, A. G., Jia, F.

    Physics of Fluids 38(9), 091905

    A network trained on twelve patient-specific simulations predicts wall shear stress in healthy, grafted and diseased coronary arteries.

    DOICode

  2. Published2025

    Fluid–structure interaction analysis of pulsatile flow in arterial aneurysms with physics-informed neural networks and computational fluid dynamics

    Ur Rehman, M. A., Ekici, Ö., Farooq, M. A., Butt, K., Ajao-Olarinoye, M., Wang, Z., Liu, H.

    Physics of Fluids 37(3), 031913 · Open access, CC BY 4.0 · Cited 24 times (OpenAlex, November 2026)

    Simulation of six vascular models, four with Marfan aneurysm and two healthy, combined with physics-informed networks that recover pressure, velocity and wall shear stress at lower cost.

    DOICode

  3. Published2025

    A hybrid physics-informed neural network–SEIRD model for forecasting COVID-19 intensive care unit demand in England

    Ajao-Olarinoye, M., Palade, V., He, F., Wark, P. A., Mukandavire, Z., Mousavi, S.

    Recent Advances in Deep Learning Applications, pp. 231–256. Chapman & Hall/CRC (Taylor & Francis)

    Joins an epidemic compartment model to a physics-informed network to estimate intensive care demand in England.

    DOI

  4. Line chart of actual ventilator bed use against single-step model predictions from January to March 2023.

    Published2023

    Deep learning based forecasting of COVID-19 hospitalisation in England: a comparative analysis

    Ajao-Olarinoye, M., Palade, V., Mousavi, S., He, F., Wark, P. A.

    IEEE ICMLA 2023, pp. 1344–1349 · Conference paper and poster

    A like-for-like comparison of deep learning baselines for hospital admissions across England.

    DOICode

Accepted and in press

  1. AcceptedNew2026

    Physics-informed multi-quantile forecasting for risk-averse NHS critical-care surge allocation

    Ajao-Olarinoye, M., Babatunde, A., Sadeghimanesh, A., He, F., England, M.

    UKCI 2026, Coventry. Springer proceedings, in press · Oral presentation, 10 September 2026

    A physics-informed model that outputs uncertainty ranges feeds a risk-averse optimiser. Unmet demand stays close to simple rules while far fewer patients are moved.

    Code

  2. Double-bulge aneurysms of 4, 5 and 6 cm coloured by wall shear stress, with CFD results above and PINN results below.

    AcceptedNew2026

    Modeling of double descending thoracic aortic aneurysms using computational fluid dynamics (CFD) and residual-based physics-informed neural networks (ResNets-PINNs)

    Fatima, N., Hamza, M., Farooq, M. A., Ajao-Olarinoye, M.

    Pramana – Journal of Physics, in press · Accepted 2 March 2026

    A residual-network model for double-bulge aneurysms of 4, 5 and 6 cm, trained on computed wall data and compared with the solver.

    Code

Preprints

  1. PreprintNew2026

    Adaptive weather-aware home energy management through large language model-based appliance scheduling

    Jonah, S., Moses, Q., Babatunde, A., Ajao-Olarinoye, M., Bammeke, D.

    SSRN preprint · Earlier version on arXiv as “LLMs for agentic home energy management”

    An agent turns everyday requests into schedules for several household loads, using prices, weather-based solar estimates and household limits. Five language-model controllers are compared with an optimisation benchmark.

    SSRNarXiv

  2. Pipeline diagram: document pages are ingested as text passages, a knowledge graph and figures and tables. Each is retrieved for a question, fused late into one prompt and passed to a swappable generator.

    PreprintNew2026

    Controlled evaluation of graph and multimodal augmentation in RAG for document question answering

    Jonah, S., Akinseloyin, O., Ajao-Olarinoye, M., Babatunde, A.

    arXiv:2607.16604

    Tests when graph and multimodal retrieval really help, with controls for text leakage.

    arXiv

  3. Preprint2025

    MSAGAT-Net: multi-scale temporal adaptive graph attention for efficient spatiotemporal epidemic forecasting

    Ajao-Olarinoye, M., Palade, V., He, F., Wark, P. A., Mousavi, S., Mukandavire, Z.

    SSRN preprint · Also under review at a journal

    A graph attention network that learns how regions influence each other over several time scales, for regional multi-step prediction.

    SSRNCode

Under review and submitted

  1. Under review2026

    Multiphysics fluid–structure interaction modeling and physics-informed neural network surrogate for morphology-dependent stress localization in thoracoabdominal aneurysms

    Ur Rehman, M. A., Temimi, H., Ajao-Olarinoye, M., Laadhari, A., Riahi, M. K., Kissami, I.

    Manuscript under review

    A fast neural surrogate for blood flow and wall stress across six thoracoabdominal aneurysm shapes.

  2. Submitted2026

    Comparative analysis of numerical schemes and physics-informed neural networks for two-dimensional dam-break problems with source terms under different initial dam breaking conditions

    Zahra, Ajao-Olarinoye, M., Farooq, M. A., Khan, F.

    Manuscript submitted

    A like-for-like comparison of classical solvers and physics-informed networks across six dam-break starting conditions.

In progress

  1. In progress

    On the orthogonality penalty in spectrally parameterised physics-informed neural networks

  2. In progress

    Hemodynamics of idealised descending thoracic aortic aneurysms and recovery of pressure and wall shear stress from sparse velocity

Talks and awards

  1. 20 Oct 2026
    Invited talk, Centre for Doctoral Training in AI launch event, Coventry UniversityFrom Student to Success: An AI Alumni Journey
    Invited talk
  2. 29–30 Apr 2026
    Invited speaker, Future Algorithms 2026 (3rd edition, online)Algorithmic training strategies for physics-informed neural network wall shear stress prediction in cardiovascular hemodynamics
    Invited talk
  3. 12 Nov 2025
    Research seminar, Centre for Computational Science and Mathematical Modelling, Coventry UniversityData-Driven Algorithms for Forecasting Critical-Care Demand during the COVID-19 Pandemic
    Seminar
  4. 11–12 Jun 2025
    First Prize, poster competition, Future Algorithms 2025 (2nd edition, online)Physics-informed neural networks for fluid–structure interaction analysis of arterial aneurysms
    Prize
  5. Jun 2024
    Best Reviewer Award, CSMM Conference, Coventry UniversityFor reviewing conference papers and giving high-quality feedback to authors
    Award
  6. 7–8 May 2024
    Talk, CSMM 2nd Annual Conference, CoventryPhysics-informed neural networks for modelling infectious disease dynamics: a case study of COVID-19 in England
    Talk
  7. 2023
    Conference presentation, CSMM 1st Annual Conference, CoventryComputational science, mathematical modelling and AI applications
    Talk
  8. 26–27 Apr 2022
    Poster, Doctoral Capability and Development Conference (DCAD22), Coventry UniversityAllocating medical resources (vaccine) to control a pandemic
    Poster
  9. May 2021
    PhD Studentship Award, Coventry UniversityFully funded studentship at CSMM to research resource allocation during pandemics
    Award
  10. Oct 2017
    Third Place, Community Development Project, KWASU Entrepreneurship and Innovation ChallengeLed a project on youth empowerment through technology
    Award

Teaching

Lecturer in Computing, QAHE, BirminghamProgramming, data science and machine learning through partnerships with Ulster University and Southampton Solent University. Teaching materials, assessment and curriculum development.
Feb 2025 – present
Dissertation supervisionMore than 10 MSc dissertations at Ulster University, supporting research design, methodology and thesis completion.
2025 – present
Graduate Teaching Assistant (part-time), The Futureworks, Coventry University GroupTutorials and workshops in information retrieval, big data analytics, deep learning and computer vision; assessment feedback and individual academic support.
Apr 2021 – Jan 2025
Volunteer lecturer, CodeYourFutureTeaching software development to learners from under-represented backgrounds. Remote.
Dec 2020 – present
Software Engineering Team Lead, AI Literacy teamAfrican Institute for Artificial Intelligence (AI4AI). Volunteer, remote.
Jan 2024 – present

CV

Download the full CV (PDF)

Research and industry positions

Research Fellow, EVOLVE AIStrategic Planning and Insight Office, Coventry University.
Sep 2026 – present
Research Fellow, HOPE-MOVECentre for Computational Science and Mathematical Modelling, Coventry University. Innovate UK-funded digital health project.
Jan 2026 – Aug 2026
Software DeveloperTammy Tech, Lagos, Nigeria. Web and mobile applications with JavaScript and Flutter.
Mar 2020 – Apr 2021
Business Systems AnalystVerificar Compliance, Lagos, Nigeria. Business analysis, dashboards and reporting.
Sep 2018 – Jan 2020

Education

PhD, Computational Science and Mathematical ModellingCoventry University. Thesis: “Data-Driven Algorithms for Forecasting Critical Care Demand during the COVID-19 Pandemic”.
May 2021 – Mar 2025
MSc Data Science and Computational Intelligence, DistinctionCoventry University. Dissertation: “Plant Disease Detection: A Cross-Platform Mobile Application Using Artificial Neural Networks”.
Jan 2020 – Jan 2021
BEng Electrical and Electronics Engineering, Second Class HonoursKwara State University (KWASU), Malete, Nigeria. Project: designed and built a low-power FM transmitter with a 2 km range.
Sep 2011 – Jul 2016

Memberships and service

Programme Committee member, 23rd IEEE ICMLA
2024
Associate Fellow of the Higher Education Academy (AFHEA)
2024 – present
Royal Statistical Society e-Student Member
2024 – present
IEEE Student Member
2021 – present
Chartered Management Institute (CMI)Level 7 Certificate in Strategic Management and Leadership; Level 7 Award in Professional Consulting.
2021

Contact

For research collaboration, speaking and peer review, send me a message here.

Privacy notice

This site is run by Dr Michael Ajao-Olarinoye, Coventry, UK. If you use this form I receive your name, email address, the type of enquiry and your message, and I use them only to reply. This is based on your consent, given by ticking the box, which you can withdraw at any time.

The form is processed by Formspree, based in the United States, so your information may be transferred outside the UK. The site is hosted on GitHub Pages, which may log technical information such as your IP address.

There are no analytics, advertising or tracking cookies. If you choose a colour theme, that one choice is saved in your browser. Fonts are served from this site.

I keep messages for as long as I need to reply, and normally for up to 12 months afterwards, then delete them. You can ask me to show, correct or delete what I hold about you using this form. You can complain to the Information Commissioner’s Office.

What is this about?
Up to 5,000 characters. Please leave out passwords and confidential documents.