Browse our student profiles here. Click their names to navigate to their profile websites. Cohort 2024Name Research ProjectProfileCameron Barker End-to-end systems for efficient machine learning After graduating with a degree in Computer and Electronics Engineering, I spent a year in industry, researching and developing hardware accelerators for Spiking Neural Networks and Large Language Models. read more This continued from research experience I gained during a summer internship and my bachelor's dissertation on neuromorphic algorithms. I am currently interested in making Machine Learning and Artificial Intelligence applications more efficient and deployable on edge devices. In order to achieve this, my research focuses on novel model design and co-optimization across the whole Algorithm/Software/Hardware stack. Andras Szecsenyi Efficient Large Language Models I completed my undergraduate degree at the University of Manchester's BSc (Hons) in Artificial Intelligence with Industrial Experience between 2020 and 2024. read more During my time there I did my honours project about retrieval augmented generation with large language models for question answering. After my graduation I took on a casual research associate position to turn this project into a policy support system for the university. As part of my programme, I had a year of industrial placement at Recourse AI, a conversational AI start-up for medical and pharmaceutical institutions with the aim of training enhancement. Here I was a key part of the AI research team working with knowledge graph machine learning systems and natural language processing. I worked here during my placement year and afterwards in my final year part-time. Benjamin Sanati Lifelong, Continual Learning Agents I completed my master's in electronic engineering with artificial Intelligence at the University of Southampton, where I focused on machine learning, deep learning, and reinforcement learning. read more Over the last few years, I have worked on multiple academic and industry-related research projects. For my bachelor's thesis, I investigated the trade-off between classification accuracy, classification specificity and latency in early-exiting dynamic DNNs. Subsequently, I became a UG research scholar, exploring efficient model compression methods for computer vision on embedded devices. For my master's project, I worked in a team to develop a mobile application designed for an industry client, requiring me to curate a custom dataset and fine-tune pre-trained object detection and image classification models. After my master's, I worked as a Senior Research Assistant in Machine Learning at the University of Southampton, working on an Innovate UK-funded project to develop a mobile 3D scanning and SLAM system for automated vehicle inspection. My PhD research focuses on zero-shot generalisation and continual learning in multi-agent reinforcement learning. Specifically, I aim to develop algorithms that learn policies invariant to domain shift, particularly in open-ended learning scenarios. Maksymilian Kret Democratising Compilers for Machine Learning Systems I have recently graduated with a BSc (Hons) in Computer Science from The University of Edinburgh. read more During my degree I have gained particular interest in the fields of Compiler design & optimization, Programming language theory & design as well as Database systems architecture and Software testing. My dissertation centered on the topic of Machine Learning Systems, particularly In-Database Machine Learning. My project involved extending three In-Database ML systems, and providing a testing framework to more easily visualize and compare their results, complete with a GUI. Throughout my degree I had the opportunity to further explore my interests and tutor several different introductory courses in Programming Language Theory, Compilers, Software Engineering & Robotics. Additionally, during the summer after my final year I completed a research internship, where I used Scala to explore a more declarative approach to Compiler design, based on a state of art Compiler framework MLIR. Matej Sandor Enhancing Fault Tolerance and Efficiency in Large-Scale Machine Learning Systems I am currently pursuing a PhD in Machine Learning Systems at the University of Edinburgh. My work aims to design and optimize infrastructure that supports large-scale machine learning applications, ensuring efficient performance and fault tolerance across distributed systems. read more As part of my undergraduate degree, I completed a thesis that explored techniques for optimizing large language model inference in serverless architectures, which deepened my interest in scalable machine learning systems. In parallel, I gained substantial industry experience through internships at Google, where I contributed to projects like YouTubeâs Content Management System, Memberships, and Rollouts, working on large-scale systems. My role at Assetario also developed my expertise in back-end development for cloud-based solutions, where I focused on enhancing automation, system monitoring, and performance optimization. Rachel Somerset David vs. Goliath: An Array of Raspberry Pi's as a replacement to Servers As a PhD student specialising in machine learning systems, my research is centred on optimising the integration of machine learning algorithms within distributed systems, and the potential of deploying machine learning inference on the edge. read more I earned my undergraduate degree with honours in Computer Science and Mathematics from the University of Edinburgh, where I developed a strong fascination for computer systems and architecture research. Notably I was able to contribute to research into developing fair data structures for stronger performance isolation guarantees within the operating system and with potential wider applications during this period. I also briefly worked as a research assistant on a project looking at the fairness of atomic instructions. Between my third and final year I completed a software engineering internship at Leonardo during which I designed and implemented a dependency network within one of their products to improve the integration of new features. Stephan Kostov Autonomous Adaptive Edge-Distributed ML My professional journey has been rooted and driven by my passion for problem solving. read more Beginning as a Mechanical Engineer, I transitioned towards my deeper interests in Maths and Computation, ultimately towards a researcher in the field of Machine Learning (ML). I hold a Bachelor's degree in Mechanical Engineering, during which I completed a research internship at Hong Kong Polytechnic University under the supervision of Dr. Hui Tang. This foundational problem-solving experience allowed me to pursue my interest in ML through a Master's degree in Data Analytics. Leveraging my engineering and IT experience, I entered industry as a Software Engineer, at a leading media company Sky. In my free time I have developed various ML applications, most noteworthy is a cooking recipe generator which uses data of food's molecular compounds, including creating a dataset, and training a custom transformer model architecture. I am now wholeheartedly pleased and excited to be beginning my research directly in the field of ML. My project is in distributed ML systems; specifically models which adapt to their local environment, and collaboratively share these adaptations with each other. Along with my professional life, I have a variety of hobbies including basketball (previously playing in a national 1st division team), rock climbing, skiing, travel, photography, piano, baking, reading, and dungeons and dragons. Kejiang (Brian) Qian Trustworthy Intelligent Agent System My academic journey spans multiple disciplines. Starting with an undergraduate and graduate background in civil engineering and urban informatics, I built a solid theoretical foundation in leveraging engineering and machine learning for practical decision-making applications. read more My subsequent 2-year computer science research at King's College London specialized in multi-agent reinforcement learning (MARL) for urban renewal to balance public interests of urban development and diverse demands of stakeholders. During my time at the MIT City Science Lab @ Shanghai, I was captivated by the challenge of integrating human factors and collective intelligence into advanced technologies, such as smart contract, multi-agent simulation, urban data mining, and human computer interaction. This integration aimed at developing efficient incentive mechanisms and fostering democracy and algorithm fairness. My contributions primarily focused on designing voting mechanisms and utilizing multi-agent systems to simulate the collective decision-making process for urban planning and sustainable transport. Anton Lydike Making Machine Learning Accelerators Easier to Use I am a first-year CDT student looking into program synthesis for accelerators. read more I did my undergrad in Germany, writing my thesis about a RISC-V emulator and a small operating system kernel that was able to run on that emulator. From there I went to work as a researcher here in Edinburgh, looking at HPC compilers and compilers for bespoke RISC-V ISA extensions and accelerators, with papers published and under review in these fields. All my work in compilers was built in xDSL/MLIR, a framework by the LLVM project focused on extending LLVM-IR to be a multi-level IR. For my PhD I want to focus on program synthesis, building on my experience in HPC and ISA extensions to find novel solutions that allow us to match generic algorithms to hyper-specific accelerator APIs without sacrificing performance. Guy Frankel Using Programming by Example to facilitate online, continual, evolution of Domain Specific Languages After completing my BSc in Biochemistry, I pursued a master's in computational science at the University of Amsterdam, with my thesis research conducted at the Weizmann Institute of Science's faculty of complex systems. read more My work focused on developing computational models to investigate ant foraging behaviour. I then began a PhD in theoretical biology at the Weizmann Institute, within the faculty of computer science, where I explored non-traditional representations of evolution, focusing on tiling, emergence, and graph neural networks. For personal reasons, I left the PhD program after two years. Next, I worked as a research assistant at the Technion Institute of Technology in the field of human-computer interaction. My research involved integrating program synthesis into developer tools, integrating programming-by-example into the VS Code debug protocol. In my PhD at the University of Edinburgh, my research focuses on combating combinatorial explosion that is inherent in the search phase of program synthesis. This article was published on 2024-12-05
