Loading…
Beyond language — across health, biology, code, financial markets and the physical world.
Foundation models are expanding beyond natural language. Across healthcare, biology, software, music, finance and the physical world, valuable information is expressed through clinical events, physiological signals, genomic sequences, molecular structures, source code, mathematical expressions, acoustic patterns, market data and physical trajectories.
This programme explores how foundation-model principles change across these different data languages.
Hover a domain to explore.
A unified journey
From shared scientific foundations to the frontiers of domain-specific intelligence.
Curriculum structure
Seven connected parts, from foundations to the physical world.
Programme focus
70% research · 30% engineering and applications.
How foundation models learn across sequences, signals, graphs and physical systems.
Representation learning, transformers, emerging architectures, scaling laws, multimodality and transfer.
How models learn from longitudinal and multimodal human data.
Clinical records, imaging, wearables, biosignals, patient trajectories, personalisation and clinical validity.
How models connect sequence, structure, function and experiment.
Genomics, proteins, molecules, chemistry, materials, virtual cells, and autonomous experimentation.
How models reason within executable and verifiable languages.
Code models, program synthesis, mathematical reasoning, theorem proving, verification and RL with verifiable feedback.
How models learn from sound, structure and composition over time.
Audio representation, music modelling, generation, separation, restoration and multimodal audio intelligence.
How models learn in adaptive and non-stationary environments.
Time-series models, limit order books, forecasting, regime change, market simulation and economic agents.
How models connect perception, prediction and action.
Earth & climate models, world models, robotics, industrial systems, digital twins and physical intelligence.
Who it's for
Researchers, engineers, scientists and technical leaders across health, biology, chemistry, software, finance, audio, robotics and the physical sciences. Participants should have a working understanding of machine learning.
What participants will gain
Natural language is only one of the many languages through which the world can be understood.
/ OxML 2027 — Oxford
Leave your email and we’ll write when applications open for the three 2027 modules.