Ineffable Intelligence
The Team
Four researchers and engineers in London. Each owns a strand of our programme. Collectively we work on the three pillars of unsupervised AI.
David Silver
Co-Founder & CEO
David's background is in machine learning research and the study of unsupervised methods. Before founding Ineffable Intelligence in 2021, he worked at a small AI research group, where he spent several years investigating why self-supervised objectives — objectives that derive training signal from the data itself rather than from human labels — were producing more transferable representations than their supervised counterparts in certain low-annotation regimes. That observation became the founding question for this laboratory. He leads the overall research direction and owns the self-supervised representation learning strand.
Amara Osei
Research Scientist
Amara completed her doctoral research at a UK university, where her thesis examined contrastive and non-contrastive self-supervised objectives and the geometric structure of the embedding spaces they produce. She joined Ineffable Intelligence in 2022. Her current focus is on probing methods for evaluating representation quality without access to labelled downstream tasks — a necessary precondition for measuring progress in the absence of annotation.
Priya Narayanan
Research Engineer
Priya specialises in distributed training infrastructure for large unsupervised models. Before joining Ineffable Intelligence, she worked at a London-based AI company building and maintaining PyTorch distributed training systems for foundation model research. At Ineffable Intelligence she owns the compute infrastructure and training pipeline that supports all experimental work — ensuring that research ideas can be tested quickly and at sufficient scale to be informative.
Tobias Wenzel
Research Scientist
Tobias works on continual learning and reward-free exploration. His background spans cognitive science and machine learning — he is interested in what the organisation of biological memory, particularly the distinction between episodic and semantic memory systems, suggests about how artificial systems might resist catastrophic forgetting. He joined the group in 2023 and leads both the reward-free exploration and continual learning threads.
Interested in working on these problems with us?
See Open Positions