Education
- MPhil Advanced Computer Science, University of Cambridge
- October 2024 – Present
- MEng Computer Science, University College London
- September 2020 – June 2024
- First Class Honours (GPA 4.0 equivalent). Average grade: 80%.
- Undergraduate dissertation on Language Model Adaptation for Low-Resource African Languages [pdf | code].
Research experience
- Student Research Assistant, Natural Language Processing Research Group, Department of Computer Science, UCL
- October 2023 – Present
- London, UK
- Data-Efficient Task Unlearning in Language Models
- Applied state-of-the-art optimisation-based machine unlearning methods to improve safety of a 1.3B parameter LLM.
- Proposed a data-efficient method for unlearning broad capabilities (in contrast to removing specific information present in the training dataset) involving as little as 32 samples of undesired prompt-responses and neutral sample pairs.
- Currently, scaling up the experiments to 8B parameter models and other tasks.
- BritLLM
- Experimented with several hardware and software configurations for 3D parallelism in a multi-GPU distributed environment (Megatron Core, Accelerate DeepSpeed, FSDP, CUDA).
- Our work is publicly available here.
- Summer Research Internship, Department of Computer Science, UCL
- August 2023 – October 2023
- London, UK
- Conducted research on multilinguality of Large Language Models. Assessed the impact of tokenisation algorithms on the performance of multilingual transformer language models.
Work experience
- Software Engineering Intern, gLinux, Google
- June 2023 – September 2023
- Munich, Germany
- Designed and automated configuration generation for internal software end-to-end release pipelines.
- Automated the setup of essential permission groups and software deployments to package repositories using the build system and internal CI tools.
- Technologies used involved Golang, Bazel with a significant share of several Google-specific internal projects and tools.
- Extreme Blue Intern, IBM
- June 2022 – September 2022
- Hursley, UK
- Created a proof-of-concept of an immersive AR experience, focused on , supporting achieving client’s goal of 45% circular transactions.
- Developed a REST API connecting AR application with client’s website and a product database.
- Technologies used included Microsoft HoloLens, C#, and Unity, in the AR platform, and Python Flask, JS, Selenium, and mongoDB in the middleware.
- Summer Studentship Intern, UCL
- July 2021 – August 2021
- London, UK
- Covered machine learning fundamentals, text pre-processing, NLP classification models and word embeddings.
Summer schools
- CMMRS 2024, Pre-doctoral Research School in Computer Science
- August 2024
- SaarbrĂĽcken, Germany
- EEML 2024, Eastern European Machine Learning Summer School
- July 2024
- Novi Sad (Нови Сад), Serbia
- Presented our preliminary results on Data-Efficient Task Unlearning in Language Models during the poster session.
Teaching
Service and leadership
- Co-lead of the Programming Tutor Initiative, UCL
- August 2022 – April 2024
- Co-organised a hiring process and managed the tutoring scheme at the university for more than 200 tutees and 20 tutors.
- In free time, built a cool website providing information and additional resources to the computer science undergraduates in the department.
- Head of Development, UCL AI Society
- August 2023 – June 2024
- Gave a guest talk on Natural Language Processing in the ML Tutorial Series, as well as appeared during the Introduction to Python and the ClimateHack.AI 2023 Workshop!
- Head of IT, UCL Leaders Conference
- October 2022 – January 2023
- Coordinated the essential technical support for conference speakers and attendees during the event.
