London, England, United Kingdom
I am currently a Machine Learning Research Engineer, working on challenging problems at DeepMind. I am also passionate about increasing diversity in the field, and strengthening participation of Africans in particular, through my involvement in the steering committee of the Deep Learning Indaba. I was previously a full-stack analyst developer and am also a Financial Risk Manager - Certified by the Global Association of Risk Professionals
I currently work as a Staff Research Engineer in the AI for Educaation research team at Google DeepMind. My role allows me the flexibility of working on cutting-edge scientific research, sound software engineering and everything in-between. Most recently my work involves fine-tuning of large language models to build an AI Tutor called LearnLM. I've previously worked on large scale multi-agent RL projects, COVID modelling and online learning, amongst other unpublished projects. Achievements Co-author of two technical reports on LearnLM - an AI Tutor based on Gemini. (https://arxiv.org/abs/2412.16429 ; https://arxiv.org/abs/2407.12687) Co-author of the paper "Learning few-shot imitation as cultural transmission" published in Nature Communications (https://www.nature.com/articles/s41467-023-42875-2) Co-author of the paper "Human-Timescale Adaptation in an Open-Ended Task Space" (https://arxiv.org/abs/2301.07608) Published the DELVE Global Covid-19 Dataset while on secondment at the University of Oxford Department of Statstics (https://rs-delve.github.io/data_software/global-dataset.html) Co-authored predictive models and visualisations of the development of COVID-19 through the UK (https://localcovid.info/) Co-author of the paper "Online Learning with Gated Linear Networks" (https://arxiv.org/abs/1712.01897)
The Deep Learning Indaba is a movement to strengthen machine learning applications and research on the African continent. It grew from humble roots in Johannesburg, to the largest annual gathering of AI researchers on the continent along with multiple programmes to grow the community over the years including 36 regional events, a mentorship programme and research grants. I am on the executive board and was a co-general chair of the Deep Learning Indaba 2023 held in Accra, Ghana, which hosted 800 of Africa's top AI talent for a week-long event of learning, research, exchange, ideation, and debate around the state of the art in machine learning and artificial intelligence. Achievements: 2023: Co-general chair of the 2023 Indaba, hosting 800 people in Accra, Ghana. Responsibilities included management of a half a million dollar budget and overseeing and managing a team of 60 organisers. Chair of Baobab committee, a custom event management system that handles applications and reviews, registration and communication with over 2000 applicants. 2022: Chair of the applications and selection committee, responsible for identifying the key community members across the continent that would act as force multipliers of the Indaba's investment. Strategic allocation of travel grants to support the community. Chair of the Baobab committee. 2021: Started an innovative new "transactional" mentorship programme. Over 400 mentees have benefitted from this programme since. * The Indaba took a COVID break in 2020 and 2021 * 2019: Chair of applications and selection and Baobab committees. Won the Umuntu prize for services to the African AI community. Gave lecture on Recurrent Neural Networks. 2018: Worked on applications & selection, co-authored practical notebook material; Lectured session on mathematics for machine learning; Organised "Machine Learning in Production" parallel session 2017: Co-authored practical notebook materials; head tutor.
Allan Gray is an investment management firm, managing over $35bn on behalf of institutional and retail clients including the largest unit trust fund in South Africa. I worked as a software developer in a team responsible for delivering IT solutions to Trading, the Investment Team and Client Services in the institutional business. I was involved in all stages of the software development life cycle from analysis and architecture to development and support. I was away on a study sabbatical from September 2015 to September 2016. Achievements Successfully developed and launched a bespoke Investment Research system to streamline analysts’ processes. I worked as the sole developer and did the majority of the analysis for this system. It is currently being used daily and receives positive feedback from users. Introduced WPF and user experience design principles to the development team. Successfully architected a web framework based on the MEAN stack for use in Institutional IT. Key Projects Investment Research System: This system was built on top of a custom generic calculation engine and integrated financial reporting and market data, handled custom reporting and managed document storage and retrieval through interaction with Sharepoint. Technologies included C#.NET, WPF, WinForms, Devexpress, SQL Server and Sharepoint. Regression Test Automation: A custom system that stores and automatically executes regression test cases for the hundreds of reports in the institutional business was built using the MEAN stack (MongoDB, Express, Angular JS, and Node JS). The system interacted with Active Directory and SQL Server and communicated with a distributed .NET service using RabbitMQ messaging. Invoicing: An easy-to-use WYSIWYG Invoicing application used by Finance and Portfolio Administration. Technologies included C#.NET, WPF and SQL Server.
Absa Capital is the corporate and investment banking division of Absa Bank Limited (owned by Barclays). I joined Absa Capital on their highly competitive graduate programme and was placed in the Front Office Pricing and Risk development team where I worked on systems for Trading, Market Risk and Portfolio Management (credit risk). I eventually became one of the lead developers on the bank's End-of-Day Profit and Loss (PnL) Explain system for Middle Office. I was heavily involved in all stages of the software development life-cycle including analysis, architecture, development and support. Achievements Successfully motivated an entirely new, system oriented architecture for the PnL system to the bank’s enterprise architecture committee. This system is still running to this day. Selected to represent the team on the enterprise architecture committee. Obtained A performance ratings in all years, including when evaluated against the highly competitive graduate cohort. Key Projects Middle Office PnL: This system needed to process PnL for the entire secondary markets trading desk in a short timespan after close of trading, making performance crucial. It was built using Microsoft’s C#.NET, SQL Server, WPF and WCF. It communicated over Solace, a high speed messaging bus and interacted with the bank’s Front Office system (SunGard Front Arena) using a Python API. It also leveraged a bespoke derivatives valuation system called CRE (Core Risk Engine). Various Excel Upload Sheets for Traders: Various derivative pricing and price upload tools were built for Traders using VBA macros and Python for interaction with the Front Office System.
Worked as a Computer Science 2 Tutor during the academic year.