is a team lead and a senior researcher in machine learning and clinical artificial intelligence.

is a DPhil student in the Translational Neuroscience and Dementia Research group. Shuhan's work involves using iPSCs-derived neurons and astrocytes as a research model to study their crosstalk and how this plays an important role in sporadic Alzheimer’s disease, aiming to identify potential therapeutic targets related to APOE4 pathologies.

is a DPhil student in the Translational Neuroscience and Dementia Research group. Itzel's research aims to robustly identify phenotypic signatures attributable to different gene variants and knock outs involved in the Triggering receptor expressed on myeloid cells 2 (TREM2) pathway in human microglia within an isogenic background. This will allow for the identification of TREM2 pathway modulators and their phenotypic effects, which ultimately will provide a drug discovery platform to therapeutically target the TREM2 signalling pathway.

is a DPhil student in the Department of Physiology, Anatomy and Genetics, University of Oxford. Sarah aims to explore the genetic risk factors in Parkinson's disease (PD). Sarah's research focuses on PD GWA-risk genes using a CRISPRi-based screen. By correlating subcellular morphological phenotypes from high-content imaging with post-translational proteomic modifications, Sarah aims to uncover novel pathways and interactions through data mining, contributing to a deeper understanding of the mechanisms underlying sporadic PD cases.

is a PhD student in the Department of Psychiatry, University College London. Yolanda explores the association between digital biomarkers and psychosocial markers of dementia risk by using large longitudinal biomedical data and data collected from wearable devices.

has worked on the problem of identifying candidate compounds for Alzheimer's drug trials using representation learning. Claudia assessed a number of candidate compounds, aiming to shift the morphology of high-risk ApoE E4 microsglia cells to neutral-risk ApoE E3 microglia cells. Two supervised feature extraction models, binary classifier and triplet loss network, were developed and enhanced with transfer learning using MobileNetV2 architecture. Additionally, unsupervised models, were explored. To mitigate the uneven cell denity problem in high-content images of cells, a novel debiasing methods was introduced. The results of this research work demonstrated that machine learning methods have the potential in pre-clinical drug discovery.

Dissertation title: "Representation Learning for Discovering Compounds Mitigating Alzheimer's Disease Risk"

Claudia Cilleruelo has recently joind Man AHL as a quantitative research analyst.

has worked on the problem of analysing semantic trajectories in schizophrenia using the path signature approach. The study focused on developing and feasibility of using the signature-based machine learning models to enhance the precision of schizophrenia diagnostics. The developed methods enabled to capturing complex interactions within sequential data to explore cognitive differences between schizophrenia patients and healthy controls. 

Dissertation title: "Using path signature representation of semantic spaces to distinguish healthy volunteers versus patients with schizophrenia"

has worked on the problem of automating and improving the accuracy of microglial phenotype classification in brain tissue images. By applying a range of computer vision machine learning models (such as YOLO and Fast R-CNN), Qiaochu has enhanced the process of identifying microglial phenotypes, which traditionally relied on subjective and time-consuming shape analysis based on human expertise. Fast R-CNN analyses specific image regions for accuracy, while YOLOv8 scans the entire image to predict bounding boxes and class probabilities efficiently. This approach not only simplifies classification and reduces bias but also allows for a detailed examination of microglial diversity, advancing our understanding of their role in central nervous system health and disease, and opening new possibilities for targeted therapeutic interventions.

Dissertation title: "Using Deep Learning to Classify Subtypes of Microglial Cells in Humans"

has worked on the problem of multiple change-points detection, which involves identifying moments when the distribution of a time signal significantly changes. Rik used latent stochastic differential equations (latent-SDEs) to model the underlying data-generating mechanisms. A change point was then identified when the difference between the solutions of the trained latent-SDE and the observed data increases suddenly. Rik's approach was applicable to both offline and online change-points detection and was compared to the UCR time series archive, comprising 120 datasets, and showed being competitive against four common benchmark algorithms for change-point detection.

Dissertation title: "Neural Stochastic Differential Equations for Multiple Change Points Detection"

Rik is now a DPhil student in the Department of Statistics, University of Oxford.

has worked on the problem of representation learning for dynamic graphs, where evolving node features and changing network structures create complex temporal dynamics. Torben extended the Graph Neural Controlled Differential Equations (Graph Neural CDEs) by incorporating permutation equivariance to ensure consistent model predictions under node permutations. This new method, Permutation Equivariant Graph Neural CDEs was aimed to address the lack of theoretical guarantees for robust performance on graphs. Evaluating this approach on diverse dynamical systems and a molecular dynamics task, Torben demonstrated its effectiveness and provided insights into its learning behavior.

Dissertation title: "Equivariant Graph Neural Controlled Differential Equations"

Torben is now a PhD student in Machine Learning and Artificial Intelligence at the Heidelberg Institute for Theoretical Studies.

has been a DPhil students in Health Data Science at the Big Data Institute and the Department of Psychiatry, University of Oxford. Niall has a background in experimental psychology, with a Bachelors from the University of Plymouth, followed by a Masters by Research at the University of Bristol. Niall's overarching interests focus on the application and uptake of machine and deep learning in the realm of mental health, with the aid of big data and wearable devices. 

Niall works on Natural Language Processing, prompt ingineering and explainable methods for clinical decision support with free-text data from UK-CRIS and MIMIC. He is developing method to generate abstractive salient reasons derived from texts to underpin the decisions made by models. 

was a postdoctoral researcher at the Department of Psychiatry, University of Oxford. Dr Liu's expertise is in computer vision, data engineering and deep learning for personalised medicine. He works on the development of neural network frameworks for fluorescent microscopy and cellular phenotyping and personalised recommendations of medications in depression and dementia. Dr Liu holds PhD in Computer Science from the University of Essex.

Now, Dr Liu is a Lecturer in Data Science at the School of Engineering Mathematics and Technology, University of Bristol.

has a background in Physics (BSc) and Data Science (MSc) from the University of Southampton. Zuzanna worked as a research data scientist at Akrivia Health, focusing on EHR data engineering, harmonisation, and statistical modelling. After moving to the University of Oxford as a machine learning research assistant, Zuzanna researched the "brain fog" phenomenon following COVID-19 hospitalisation, managing multimodal data pipelines and contributing to a publication in Nature Medicine journal. Currently, Zuzanna is a health data scientist with the Translational Neuroscience & Dementia Research group (Department of Psychiatry), leading on the development of novel methods for morphological profiling of single-cell high-content imaging data to support drug screening and discovery.

Zuzanna is an incoming DPhil student (October 2024) in multimodal learning for population health studies at the University of Oxford.

Yi Zhang worked on high-content image recognition, geometric deep learning (GDL) and graph neural networks (GNN). His particular interest is in the application of GDL and GNN to de novo drug discovery and phenotypic screening. Yi recieved BSc (University of Manchester) and MSc (University if Oxford) both in Mathematics . His dissertation was on matrix perturbation theory, numerical linear algebra and the theory of random walks.

Yi Zhang has been a research assistant in machine learning for Natural Language Processing and has been leading on the developing of a Large Langauge Modelling approach to long-text classification tasks using electronic health records in secondary mental health as part of the CHRONOSIG project.

Now is a PhD student in machine learning and artificial intelligence at the Technische Universiteit Delft.

has worked on the problem of identifying moments in time where the properties of a time series change abruptly, known as change-points. To address this challenge, Vivian has developed a new unsupervised method using neural stochastic differential equations (nSDEs). This approach allows the model to capture the law of a given time series, enabling it to predict the signal and effectively detect these significant shifts.

Dissertation title: "Change-Point Detection using Neural Stochastic Differential Equations"

In 2023, Vivian Kessler has joined Oepfelbaum IT Management AG as a Trading Technology Engineer.

has worked on the problem of leveraging electronic health records (EHRs) to improve patient care. Jay's research focused on applying Geometric Deep Learning (GDL) to these complex datasets, utilising graph neural networks (GNNs) to analyse patient information. The goal was to develop a support tool to enhance predictions of patient readmission by inferring outcomes from hospital readmission data, to augment clinical decisions.

Dissertation title: "Geometric Deep Learning for Clinical Decision Support"

has worked on the problem of enhancing cell profiling in the Cell Painting assay, which uses fluorescent dyes to capture cell phenotypes across multiple imaging channels. Ying aimed to apply classic deep learning methods, such as U-Nets and GANs, as well as more advanced methods of image generation, such as stabel diffusion models, in order to predict the fluorescent markers using only bright field images of cells.

Dissertation title: "Stable Diffusion Models for Artificial Cell Painting"

has worked on the problem of addressing and evaluating uncertainty in machine learning models, particularly neural networks. Louis has explored the aleatoric and epistemic uncertainties, how they manifest in models, and how to represent them to provide reliable estimates in clinical decision support tools. By applying Bayesian framework to neural networks, Louis computed uncertainty measures on various models using medical data. Louis compared a single neural network that was trained to predict the distribution of multiple diagnese to a Mixtire of Experts (MoE) architecture, where each expert neural network was train on a single specialist diagnostic area, to test the hypothesis that certain model structures yield smaller uncertainty estimates than others.

Dissertation title: "Quantifying Uncertainty in Clinical Decision Support Tools"

is currently a MMath Mathematics student at the Mathematical Institute, University of Oxford. Tianyang‘s research interests lie within natural language processing and statistical analysis and works on reducing the  self-attention complexity in Transformers architectures by using low-rank tensor approximations. The aim of Tianyang‘s work is to allow transformers-based models, in particular for NLP tasks, to capture long range dependencies in electronic health records to support clinical decision making.

Dissertation title: "Kerformer: Linear Transformer with Kernelised Self-Attention"

is currently a MMath Mathematics student at the Mathematical Institute, University of Oxford. Jacob's background is in machine learning and data science, with a particular interest in  a hospital the scalability and accuracy of machine learning systems. Jacob's research is dedicated to modelling the path of a patient over the course of a treatment period. In particular I am excited about using neural controlled differential equations to try and understand the dynamics of a care plan over time.

Dissertation title: "On the Applications of Neural Differential Equations For Clinical Decision Support"

Now is a Quantitative Researcher at Aspect Capital

is currently a MMath Mathematics student at the Mathematical Institute, University of Oxford. Haobo has a background in statistics and statistical machine learning. Haobo's research is dedicated to learning models for actionable recourse in healthcare. He uses methods of adversarial neural network learning and optimisation to derive counterfactual explanations for decision support systems.

Dissertation title: "Learning models for actionable recourse and optimization for machine learning augmented decision making"

is currently a MMath Mathematics student at the Mathematical Institute, University of Oxford. Ziming has a background in computational statistics, machine learning and Natural Language Processing. Ziming’s research is dedicated to a new approach for zero-short learning with very large pre-trained language models – the Prompt Learning (PL) paradigm. He explores whether PL can be adapted to downstream analytical tasks to support clinical decision making, derive explanations and how to optimise the learning process. 

Dissertation title: "Clinical-Prompt: a systematic analysis of prompt learning to inform clinical decision making"

Now is pursuing a MSc in Financial Mathematics at the University of Oxford.

studied towards Part C dissertation at the Mathematical Institute, University of Oxford. Yuchen's research is focused on statistical analysis and machine learning modelling for drug discovery for Alzhemer's disease. He analyses glial fluorecent cell images generated using the Cell Painting method with the aim to understand the mechanism of action of candidate drug compunds with higher accuracy than existing approaches. 

Dissertation title: "Image-based profiling and machine learning for Alzheimer's disease drug discovery using human glial cells"

Now at Squarepoint Capital.

studied towards Part C dissertation at the Mathematical Institute, University of Oxford. Piotr's research focused on the development of deep learning convolutional models for analysing fluorescent microscopy images and the development of fast screening tool for morphological changes of cells treated with various compounds.  Mr Kalinowski has been a machine learning researcher at the University of Oxford (2021-2022) leading on the development of deep learning tools for high-content image analysis and to characterise cell health phenotypes and pattern recognition in mechanism of action of candidate drug for Alzheimer’s disease.

Dissertation title: "Machine learning methods for cell feature extraction in drug discovery for Alzheimer's disease"

Now is a PhD student in artificial intelligence at the University of Heidelberg.

studied towards Part C dissertation at the Mathematical Institute, University of Oxford. Wooseok's research is focused on the interface between deep learning methods for fluorecent microscopy image segmentation and topological data analysis. He developed a novel approach to derive persistence silhouettes and landscape signatures to quantify the geometry of cells' nucleui for downstream computational tasks.

Dissertation title: "From Augmented Microscopy to Topological Transformer: New Breakthrough in Deep Learning-based Cell Image Analysis in Alzheimer's Research"

Now is a research scientist at Vuno

studied towards the MSc Mathematical Sciences dissertation at the Mathematical Institute, University of Oxford. Alec's research is focused on artificial cell painting, whereas he develops deep learning methods to translate fluorescent information from stained images to non-stained ones. His work can speed up the statistical analyses of image-derived features and help biologists to study the morphological changes in cells by using bright-field images.

Dissertation title: "Artificial Cell Painting"

Now is a Owkin working on AI for drug discovery.

studied towards Part C dissertation at the Mathematical Institute, University of Oxford. Pawel's research is focused on developing new methods for generating high-resolution fluorecent images through adversarial training.