Student profiles

23/24 academic year

Ming Liang Ang 

Ming Liang is currently pursuing his PhD in the algorithmic and theoretical foundations of machine learning. He is interested in understanding the conditions under which transfer learning effectively occurs and hopes to elucidate key design principles that can enable more effective transfer learning to take place, especially in foundation models. Prior to the Foundational AI CDT, he obtained a degree in mathematics from the National University of Singapore (NUS) and was awarded the Ho Family Prize for being the top applied mathematics student of his cohort.

Wiem Ben Rim

Wiem joined the CDT in September 2020

Kai Biegun 

Kai is primarily interested in developing Reinforcement Learning (RL) algorithms in partially observable environments. He is currently exploring ways to integrate probabilistic latent variable modelling techniques into RL algorithms to learn world models in order to plan and act under uncertainty. Prior to his PhD, Kai obtained a MSc in Machine Learning at UCL, and a MSci in Mathematics and Physics at Durham University.

Sierra Bonilla 

Neural Rendering

Sierra Bonilla is interested in neural rendering frameworks for surgical scene reconstruction. Prior to this, Sierra completed a Master's in Medical Physics & Biomedical Engineering at UCL and holds a Bachelor's in Bioengineering and Mathematics minor from the University of Washington. With four years of prior experience in academic and industrial healthcare research, Sierra is dedicated to pushing the boundaries of AI for healthcare problems and beyond. https://sierrabonilla.com/

Natalia Burton 

I'm particularly interested in neuro-inspired AI and AI safety. My research focus in the development of bio-plausible and energy efficient architectures capable of generalising and succeeding in goal-oriented tasks whilst minimising power consumption. I have a degree in Electronic Engineering and a Master's in Business Management and Engineering. Prior to joining the Foundational AI CDT, I held a number of positions in tech-driven companies, most recently as Solutions Architect helping financial institutions to leverage the power of graph analytics and contextual data in uncovering fraud and other financial risks. 

Milo Carroll 

I am currently researching generative models for human motion synthesis. My primary objective is to understand how motion data can serve as a useful foundation for controlling physically simulated characters and robots in a reinforcement learning context. Additionally, I am exploring ways to efficiently repurpose these generative policies for downstream tasks, such as incorporating vision and language. My research primarily focuses on the fields of Reinforcement Learning, Robotics, and Deep Generative Models.

Francesca Channon 

Francesca Channon is currently pursuing her PhD, focusing on the application of machine learning techniques to enhance energy efficiency in the built environment. Her research aims to address the limitations of traditional energy assessment methods by integrating supervised learning, probabilistic approaches, and physics-informed machine learning. Prior to her PhD, Francesca obtained a degree in Mathematics from Warwick University, and an MSc in Statistics, and MSc in Applied Biotechnology and Biosciences from Imperial College London. She has industry experience in sustainable investment due diligence and alternative investments at J.P. Morgan Private Bank, and consulted on ESG-related topics at Tellus Matrix. Francesca's research is supervised by Professor Tomaso Aste and Professor John Shawe-Taylor. 

George Drayson

Natural Language Processing and Generative AI

George is a PhD student at the Foundational AI CDT at University College London, supervised by Professor Lampos. George is interested in Natural Language Processing and Generative AI and their applications in healthcare. Prior to starting his PhD, he obtained a MEng in Engineering Science from the University of Oxford. See georgedrayson.com for more details.

Rares Dolga 

Rares is interested in probabilistic machine learning largely focusing on methods for efficient language generation.  Prior to his PhD, he was a quantitative developer and then an AI researcher at JP. Morgan where he worked on NLP problems in finance. Rares obtained an MEng in computer science from UCL in 2020. 

Ahmet Guzel

Ahmet is interested in developing generalist AI agents that improve their capabilities through synthetic data and exhibit creativity in open-ended environments. Before pursuing this, he completed a year-long deep learning research internship at HUAWEI R&D in Cambridge, UK, and holds an MSc in Artificial Intelligence from Leeds University. He also has extensive experience in motorsport engineering
 

Nathan Herr

Cooperative AI, Generative AI, Reasoning and Planning, Multi-Agent Systems, Reinforcement Learning

Nathan is passionate about using AI to create positive societal impacts. His research focuses on enabling effective, natural cooperation and coordination between AI agents and humans in various tasks. Nathan's research interests include enhancing complex sequential decision-making processes, especially through improved exploration in large language models (LLMs), to help AI agents navigate and solve more complex problems.
Outside of academia, Nathan enjoys staying active and embracing the outdoors, taking full advantage of the UK's rare sunny days

Hongyu Lin

Hongyu Lin is a PhD student at the CDT in Foundational AI, supervised by Professor Tomaso Aste and Professor John Shawe-Taylor. His research interests lie in statistical learning theory and causality. Prior to the PhD, he obtained a BSc in Mathematics and Physics from UCL, an MAST in Applied Mathematics from the University of Cambridge, and worked as a data scientist in the renewable energy industry. 

Varsha Ramineni 

AI Bias and Fairness, Synthetic Data 

Varsha is a PhD student and DeepMind scholar in Artificial Intelligence (AI) at UCL, supervised by Prof. Emine Yilmaz and Prof. David Barber. She is passionate about bridging the technical and societal aspects of AI systems through interdisciplinary research, with particular interest in AI bias and fairness, and synthetic data. Varsha holds an MSc in Statistical Science from Oxford University and a BSc in Mathematics from Warwick University. Before joining the PhD program, she worked in mental health research, where she applied Bayesian adaptive design techniques to evaluate a digital intervention for healthcare staff experiencing intrusive memories of traumatic events.
 

Leonie Richter 

I am interested in reward models, deception and reinforcement learning from human feedback, specifically in the context of LLMs. I would like to contribute to the problem of aligning powerful artificial models with human values.

Prior to my PhD, I completed an internship at the Center for Human-Compatible Artificial Intelligence in Berkeley and worked as a Machine Learning Engineer. Before that, I completed my MSc in Artificial Intelligence at Imperial College London.

Romy Williamson

Geometry Processing

Romy works in the intersection between classical geometry processing (differential geometry, and mesh algorithms) and modern techniques (neural networks). Prior to starting the PhD, Romy earnt a BA in Mathematics from the University of Oxford and an MSc in Computer Graphics, Vision and Imaging from UCL. During the PhD Tomy is focused mainly on the representation and processing of smooth surfaces. In another project, Romy has worked on simulation, PDEs and the Finite Element Method. Romy works in the UCL Smart Geometry Processing Group, under the supervision or Prof. Niloy Mitra. https://romyjw.github.io
 

22/23 academic year

William Bankes

I am broadly interested in studying probabilistic machine learning and how uncertainty quantification can enable machine learning models to make more intelligent decisions.

Prior to my PhD I spent two years working as a Machine Learning Engineer in Canary Wharf before completing an MSc in Computational Statistics and Machine Learning at UCL.

David Chanin

David studies mechanistic interpretability in LLMs. He is particularly interested in knowledge representation in LLM hidden activations and unsupervised interpretability techniques like Sparse Autoencoders (SAEs). Prior to starting his PhD, David worked as a software engineer, and holds degrees from Stanford University and Georgia Tech
 

Zonghao (Hudson) Chen

Hudson is supervised by François-Xavier Briol and Arthur Gretton. His research interests center around Wasserstein gradient flows, numerical integration and causal inference. Prior to his PhD, he obtained his bachelor's degree from the department of Electronic Engineering, Tsinghua University. Hudson also organizes the JumpTrading/ELLIS CSML seminar series on Computational Statistics and Machine Learning for the UCL ELLIS Unit.
 

Gbetondji Dovonon

Gbetondji is broadly interested in deep learning, causality and graph theory, with the goal of designing more reliable and interpretable supervised learning systems that explicitly build on domain knowledge. He is a student at UCL and part of the ELLIS PhD program under the supervision of Matt Kusner and Michael Bronstein.

Jianwei Liu

Reinforcement Learning, Legged robots, Lifelong learning, Neuromorphic learning 

Jianwei's current research direction involves utilising Generative ML models (e.g. diffusion models) and Neuromorphic learning for planning for legged robots. He's also interested in exploring topics such as continual learning and open-endness. Prior to starting the PhD, Jianwei obtained a MEng in Biomedical Engineering from Imperial College London, a MSc in Robotics and Computation from UCL, and worked in industry as a robotics software engineer on off-shore NDT inspection robots and construction robots.
 

Hossein A. (Saeed) Rahmani

Conversational AI, Responsible IR, Recommender Systems

Hossein (aka Saeed) is a Ph.D. student in Foundational Artificial Intelligence at UCL AI Centre working on Fairness and Bias in Conversational Systems. Prior to starting his Ph.D., he received his MSc from the University of Zanjan, where he was working on context-aware recommender systems. His research interests lie in Information Retrieval, Recommender Systems, and Responsible AI. See https://rahmanidashti.github.io/ for more details.

Stuart Shanks

Reinforcement Learning, Continual Learning, Legged Robots

Stuart is interested in the intersection of optimal control and machine learning for robotics. He is supervised by Dr Dimitrios Kanoulas and Dr Carlo Ciliberto, and is working in the Robot Perception and Learning (RPL) Lab. He is particularly interested in continual learning algorithms for use on legged robots.

Hengyi Wang

Hengyi Wang is a PhD student at CDT in Foundational AI at UCL, primarily supervised by Prof. Lourdes Agapito. His research topic is "Neural Surface Reconstruction from Video". He is particularly interested in designing learning-based algorithms for real-time surface reconstructions from visual signals. Prior to starting his PhD,  he obtained a BSc in Telecommunications Engineering with Management from the Queen Mary University of London and an MSc in Computer Graphics, Vision, and Imaging from University College London. For more information please check https://hengyiwang.github.io/

Jiayi Wang

Jiayi Wang is a first-year PhD student with broad interests in artificial intelligence. She is currently working on natural language processing and adversarial machine learning.

Lorenz Wolf

Multi Agent RL, Generative Modelling, Unsupervised Learning

Lorenz is generally interested in Multi Agent Reinforcement Learning, Generative Modelling, and Unsupervised Learning, and he is supervised by Mirco Musolesi. Prior to his PhD, Lorenz obtained his MSc in Statistics and BSc in Mathematics from Imperial College London.

21/22 academic year

Rokas Bendikas

Autonomous surgical robotics

I am a PhD student at University College London, working in the field of Surgical Robot Control and Robot Learning. My research project is focused on ‘Learning priors of high-level and low-level motion representations for autonomous robot control’. Therefore, I am interested in answering questions: ‘what perception modalities provide the optimum medium for an agent to learn’ and ‘how to learn complex behaviours that allow tackling long-horizon, multi-staged objectives in surgical setting’. During my Master’s studies, I completed my thesis at Dyson Robotics Lab, supervised by Prof. Andrew Davison. My project introduced imagination-augmented DQN, which allowed to learn optimal behavioural policy in dynamically complex environments. Previously I was a research assistant in Cardio-Electro Magnetic Research Group (CEMRG), investigating atrial fibrillation mechanisms through the means of Deep Learning and Computational Modelling. My work was supervised by Dr. Caroline Roney and Prof. Steven Niederer.

Jake (Harry) Cunningham

Research Area: Gaussian Processes and Climate Science

Jake is interested in understanding and improving uncertainty quantification in machine learning methods, focusing largely on Gaussian processes and kernel methods. He also has strong interests in applying statistical learning methods to problems in climate science. Prior to his PhD, he obtained an MEng in Engineering from the University of Oxford and MSc in Computing (Machine learning and AI) from Imperial College London.

Emilio Mcallister Fognini

Applications of AI to PDE's

Emilio is interested in using AI and machine learning to aid in the solving the equations that govern physical systems. His primary area of focus is efficiently and quickly solving general Helmholtz problems, with research interests in Inverse problems and applications of AI to scientific computing and physical systems. He obtained a Master of Mathematics from UCL.

Akbir Khan

Cooperative AI

Akbir is interested in studying techniques to ensure cooperative AI. He considers these problems primarily in the setting of multi-agent deep reinforcement learning and social dilemmas. He is advised by Tim Rocktäschel and Edward Grefenstette and affiliated with UCL DARK lab. For more information check https://www.akbir.dev

Aengus Lynch

Causal Inference

Aengus is interested in improving neural networks’ ability to generalise to unseen domains. He is advised by Ricardo Silva and associated with the Causal Machine Learning Group at UCL

Yao Lu

Natural Language Processing

Yao is interested in everything, especially human language model interaction (HLMI). The ultimate goal of his research is to create a unified communication interface between human and machine through human language.

Zak Mo

rgan

Computer Vision

Zak is a PhD student on the foundational artificial intelligence CDT at UCL, prevously having obtained his undergraduate masters in computer science at UCL. He is working under the supervision of Youngjun Cho and Sriram Subramanian in the UCL Interaction Center (UCLIC) primarily on the fields of computer vision and human-computer interaction.

Mariia (Masha) Naslidnyk

Gaussian processes and kernel methods

Prior to starting her PhD, Masha was a Machine Learning Scientist at Amazon Research in Cambridge, where she worked on Alexa question answering (2015-2019), and then on Gaussian processes for supply chain emulation (2019-2021). Masha graduated from Part III in Pure Mathematics at the University of Cambridge in 2014. Her research interests lie broadly in the topics in Gaussian processes, kernel methods. Masha sits within the Fundamentals of Statistical Machine Learning research group, advised by F-X Briol, Jeremias Knoblauch, and Carlo Ciliberto.

Daniel Augusto de Souza

Interpretable Gaussian processes for geospatial problems

Daniel is broadly interested in the intersection of Bayesian machine learning models and the needs of their practitioners in the real world and how to potentially correct this mistmatch. This includes how theoretical benefits of Bayesian modelling may translate or not into practical benefits, including supposed interpretability benefits. Before joining UCL, he obtained his MSc and BSc degrees in Computer Science at the Federal University of Ceará in Brazil

Laura Ruis

Machine Learning

Laura is interested in understanding the generalization conditions of neural networks. See http://lauraruis.com for more details.

20/21 academic year

Reuben Adams

Statistical Learning Theory, in particular PAC-Bayes

Reuben’s research is on PAC-Bayes, which is a suite of tools used to bound the performance of various machine learning algorithms. He has been working with Benjamin Guedj and John Shawe-Taylor on extending classical algorithms to the multiclass regime and other domains where it is important to track the different kinds of errors that can be made, rather than treating them all equally.
He is the host of the Steering AI podcast, where he talks with academics about their work and the risks they perceive surrounding the development of AI. He has given talks on the topic of AI Safety at several universities, taken part in two debates on whether AI may pose an existential risk in the future, and organised workshops helping people develop their own views on the future of AI

Dennis Hadjivelichkov

Self-Supervision for Robotics

Denis is working on bridging the gap between robotics and intelligence through machine learning. He is part of the Robot Perception and Learning Lab where he is currently dealing with self-supervised learning of object descriptors and affordance using cues provided by stereo vision, egomotion and kinematics.

Alex Hawkins-Hooker

Protein design, deep generative models

Alex is broadly interested in the development of machine learning methods for the study of biological molecules. His PhD research focusses on the use of deep generative models for protein design and is supervised by David Jones and Brooks Paige.

J

ean Kaddour

Causal Inference and Deep Learning

Jean works with Ricardo Silva and Matt Kusner on Causal Inference and Deep Learning. Jean obtained a MSc. in Advanced Computing at Imperial College London.

Causal Inference, Machine Learning

Oscar Key

Machine Learning

Oscar's PhD research investigates how we can make large machine learning models more computationally efficient by designing algorithms are better optimised for the underlying hardware. He applies this work to both large language models and weather forecasting. Oscar is supervised by François-Xavier Briol and Marc Deisenroth, and his background is in computer science and software engineering. For more details on the research, see his website.

Robert Kirk

Reinforcement Learning, Generalisation, Out-of-distribution Robustness, AI Safety

Robert is interested in reinforcement learning, especially generalisation and out-of-distribution robustness in reinforcement learning. He wants to understand how we can build reinforcement learning systems that generalise both capably and safely to unseen environments so that we can deploy them in the real world. He's supervised by Tim Rocktäschel and Edward Grefenstette as part of the UCL DARK Lab, and was previously a software engineer, after completing an integrated masters of mathematics and computer science at Somerville College, University of Oxford.

Yicheng Luo

Meta-learning, Probabilistic Programming, Reinforcement Learning, Deep Generative Models
 

Yicheng Luo is a second Ph.D. student at UCL AI Centre supervised by Prof. Marc Deisenroth and Prof. Edward Grefenstette. He received his MEng degree from the Department of Computing, Imperial College London. For his thesis, he worked on probabilistic program analysis with Dr. Antonio Filieri. He is interested in the design and implementation of efficient and scalable Bayesian inference methods with applications in robotics, reinforcement learning, and probabilistic programming. In particular, he is currently interested in understanding how to utilize offline reinforcement learning pretraining to achieve sample-efficient online learning.

Mirgahney Mohamed

Mirgahney H. Mohamed is a PhD student at University College London working on 3D computer vision and uncertainty estimation. He obtained his master's degree in Machine Intelligence from AMMI - African Master in Machine Intelligence at AIMS, and undergrad in Statistics and Computer Science from University of Khartoum Faculty of Mathematical Science."

Oliver Slu

mbers

Multi-Agent Reinforcement Learning

I am interested in how we can design algorithms to allow for effective agent interaction in multi-agent interactions. In particular, I work on bridging the gap between multi-agent systems and Game Theory and how we can apply the large body of Game-Theoretic techniques to interacting reinforcement learning agents. Prior to joining the CDT, I completed my BSc in Economics at the University of Warwick and also the MSc in Computational Statistics and Machine Learning at UCL.

Yuchen Zhu

Causal Machine Learning, AI robustness

Yuchen is interested in making machine learning robust through modelling causal invariances within data. She develops both theory and algorithms for unbiased causal effect inference. In her investigation, she often uses kernel methods and methods inspired by the econometrics literature. Previously, she earned her Master's degree in Machine Learning from UCL, and a Bachelor's degree in Mathematics from Cambridge. She is advised by Professor Ricardo Silva and Dr Matt Kusner.

Sicelukwanda Zwane

Safe Learning in Robotics

Sicelukwanda is interested in designing safe and robust algorithms for learning tasks in robotics settings. He is a member of the Statistical Machine Learning lab where he studies model-based reinforcement learning with Gaussian Processes and different approaches to incorporating safety constraints when learning robot tasks. He is supervised by Marc Deisenroth and Dimitrios Kanoulas.