Nando de Freitas

Few careers in AI are as unconventional as Nando de Freitas's. He was born in Zimbabwe, grew up as a refugee during the war in Mozambique, and eventually became one of the most influential machine learning researchers in the world. Before entering academia, he sold beer illegally in South African townships to survive. That journey, from the margins to Trinity College Cambridge, says a lot about the tenacity he later applied to the hardest problems in AI.

His doctoral work at Cambridge (completed in 2000) focused on Bayesian methods for neural networks: the idea that a machine can not only learn patterns but also quantify its own uncertainty about what it knows. That statistically rigorous perspective set him apart early on. But the moment that made him widely known beyond academic conferences came in 2015, when he recorded his deep learning course at Oxford and posted it to YouTube. The lectures were dense and technical, with real code in Torch, and yet millions of people watched them. For many, they were the first serious entry point into modern AI.

In 2014, while a professor at Oxford, Google acquired Dark Blue Labs, the research spinoff where he worked alongside Oxford colleagues, bringing him formally into DeepMind. Over the years he rose to VP of Research at Google DeepMind, leading teams that built generative image and audio models including Imagen 3 and Lyria. In September 2024 he joined Microsoft AI as VP, focused on large-scale multimodal AI, though his tenure there was relatively brief.

What I admire most about Nando is that he never retreated into the ivory tower. He answered student questions on Twitter, published code, recorded free lectures. He believed that open education could change who gets to do research, and he was right: part of the global explosion of AI talent over the last decade carries his fingerprints. If you use Claude today or any generative AI tool, there is a good chance someone on that team learned from his videos.