CAS Fellow appointed to UN AI Panel, Vukosi Marivate helps produce International, Evidence-Based Assessment of Artificial Intelligence
Professor Vukosi Marivate is Director of African Institute for Data Science and Artificial Intelligence (AfriDSAI) at the University of Pretoria. He leads Data Science for Social Impact, advancing AI for Africans by Africans. Marivate joined the CAS Harvard South Africa Fellowship Program (HSAFP) in 2020-2021
UNITED NATIONS – Vukosi Marivate, a Harvard South Africa Fellowship alumnus, is helping produce the work of the United Nations’ Independent International Scientific Panel on AI, following his appointment to the panel by the UN General Assembly in February 2026. The panel, the first global scientific body devoted entirely to artificial intelligence, has published its Preliminary Report: Evidence-Based Assessment of Opportunities, Risks and Impacts of Artificial Intelligence. The report examines the opportunities and risks of AI and its effects on societies around the world.
Marivate was one of 40 experts selected from more than 2,600 applicants across 140 countries. The South African computer scientist has spent his career ensuring that Africa is not merely a consumer of artificial intelligence but one of its builders. For Marivate, the UN appointment was less a sudden arrival on the international stage than recognition of work he has been doing for years.
Born in Ga-Rankuwa, north of Pretoria, Marivate’s path toward artificial intelligence began with a childhood curiosity about what happened beneath the surface of a computer. He was interested not only in what these machines could do but also in how the information fed into them could become the basis for decisions. “From a young age, I was interested in how computers actually work,” Marivate wrote in an interview with the Harvard University Center for African Studies. “This led me to think about how they ingest data and are then used to make decisions.”
That curiosity took him to the University of the Witwatersrand (South Africa), where he studied electrical engineering and began working in machine learning under Professor Tshilidzi Marwala, who later supervised his master’s degree. Marivate then travelled to Rutgers University as a Fulbright Science and Technology Fellow. There, he completed a doctorate in computer science focused on reinforcement learning, the branch of machine learning concerned with how systems learn through actions and feedback.
As his technical knowledge deepened, Marivate became increasingly interested in the human consequences of the technology. “It was during this time that I started really looking at the societal effects of data science and AI,” he said. After completing his doctorate, he began turning his attention towards language and the question of what happens to communities whose words, names, and histories remain largely illegible to machines.
That problem is particularly urgent in Africa. The continent is home to more than 2,000 languages, yet most are poorly represented in the data used to train large language models. A system may appear astonishingly intelligent in English, then become considerably less functional when confronted with isiZulu, Sepedi, or Setswana. Intelligence, it turns out, can be rather provincial.
In 2017, Marivate helped establish the Deep Learning Indaba, an annual gathering created to strengthen machine learning research and expertise across Africa. He is also a chief investigator at Masakhane, a grassroots research community developing language technologies for African languages. Its name roughly means “we build together” in isiZulu, an unusually accurate name for an organization whose community now includes more than 1,000 participants from 30 African countries.
After returning to South Africa, Marivate worked at the Council for Scientific and Industrial Research before joining the University of Pretoria in 2018. He now serves as a professor of computer science and holds the ABSA-UP Chair of Data Science. He also leads the Data Science for Social Impact research group, whose work applies data science to problems in areas including energy, public safety, and public services.
Marivate became a Harvard South Africa Fellow in the 2020–2021 academic year, in which he wanted to complement his technical experience with a better understanding of the business side of the field through the CAS Harvard South Africa Fellowship, studying Leadership Development at the Harvard Business School. He had first heard about the Harvard South Africa Fellowship Program (HSAFP) through Professor Marwala, himself a former fellow.
Established in 1979 in response to the educational exclusions of apartheid, HSAFP provides mid-career South Africans with up to a year of study at Harvard. “I had come back into academia after being away for about three and a half years and wanted to connect what I was teaching in data science and AI with being able to understand the business side as well,” he explained. “The HBS Leadership Development Program was perfect for this.”
His leadership case focused on establishing an artificial intelligence institute at the University of Pretoria, one that would allow him “to scale ideas on Pan-African AI.” It was an ambitious classroom proposal, but it did not remain in the classroom. In 2025, the university established the African Institute for Data Science and Artificial Intelligence, with Marivate as its inaugural director. Later that year, a $1 million gift from Google gave the institute significant support for African-led research, including to grow a generation of African data scientists and AI leaders, build sustainable AI research communities across Africa, build ethical AI initiatives that prioritize social good, and support the University of Pretoria’s broader strategy of research and innovation to address pressing societal issues.
That said, local does not mean isolated. Marivate’s own career stretches from Ga-Rankuwa to Johannesburg, New Jersey, Cambridge, and now the United Nations. The distinction is between participating in the world and waiting for the world to deliver ready-made solutions.
Marivate is spending his 2026 sabbatical considering how language models for African languages should be evaluated. A model that performs well on a benchmark created elsewhere may still be inaccurate, culturally clumsy, or simply useless in a particular African setting.
Asked how HSAFP shaped the work he does today, Marivate pointed to the relationships he formed at Harvard Business School and the Center for African Studies. “It was a great experience interacting both at HBS and the Harvard Center for African Studies,” he said. “The network has been invaluable.”