The City PhD student works with Number 10 Downing Street civil servants to modernise a ministerial ‘Red Box’ using a large language model (LLM).
Riad Ibadulla, a deep learning engineer and PhD student in City, University of London’s Department of Computer Science, was part of the winning Team Redux in the three-day Generative AI Hackathon for Evidence House members, held from July 26th to July 28th 2023.
The hackathon’s purpose was to generate fresh insights and prototypes to demonstrate how AI can be utilised to modernise some of the core functions of the Civil Service.
Team Redux comprised Riad and civil servants who worked on the problem of modernising the ministerial ‘Red Box’ by utilising a large language model (LLM). As a reward for developing this innovative solution, the team was invited to 10 Downing Street on September 11th 2023 to present their work.
Evidence House, led by the 10 Downing Street Data Science team (10DS), is an initiative launched in January 2023, which aims to radically upskill civil servants in data science, development and AI while delivering fresh insights and innovative solutions to government problems. It has already amassed a community of over 500 engineers, developers and analysts from across government.
The Generative AI Hackathon was hosted at City and delivered in partnership with industry leaders including Open AI, Anthropic, Microsoft and Google.
Riad possesses academic and industrial experience in the field of artificial intelligence.
He holds a BEng in Computer Systems Engineering (First Class Honours) from City, where he gained his initial interest in AI after his final year project based on speech recognition using LSTM neural networks. Later he joined the University of St Andrews and completed an MSc Artificial Intelligence, where he focused more on computer vision problems of AI with a dissertation on simplex optimisation for aerial image stitching.
After a year in industry, where he worked on the optical AI accelerators, Riad decided to proceed with a PhD in the relevant field.
His doctoral research involves optimising convolutional neural networks and using vision transformers for the optical AI accelerators.
A key contribution has been developing the 'FatNet Transformation' architecture, which was published in MDPI’s AI journal.