Postdoctoral Research Assistant in Machine Learning

We are seeking

a full-time Postdoctoral Research Assistant to join Torr Vision Group at the Department of Engineering Science, in central Oxford. This post is supported by Professor Philip Torr’s Schmidt Science AI2050 fellowship and is for two years in the first instance.What is the core problem the proposed work seeks to solve? “It is 2050. Artificial intelligence can now explain its reasoning, simulate complex societies, and help humanity learn from ten thousand years of recorded history. Historians and anthropologists work with AI collaborators that reconstruct lost evidence, test competing explanations, and reveal new insights into how cultures evolve and societies endure. This project lays the foundation for that future.” The postholder will contribute to one or more of the following strands: • Foundational research on large-scale / foundation models and agentic architectures for autonomous social-science reasoning and planning. • AI Social Scientist / AI Historian research: causal reasoning methods for reasoning over heterogeneous historical and social data (texts, maps, images, archaeological records), combining causal discovery, multimodal modelling, and agent-based simulation to produce open, reproducible tools and datasets. • Infrastructure and benchmarking for large-scale social-science simulation and validated workflows. Candidates should possess a PhD (or be near completion) in PhD in Computer Science, AI, Security, or a related field. Demonstrable expertise in foundation models / large language models, multimodal modelling, or agentic / multi-agent systems is essential. Experience or strong knowledge of causal discovery methods and/or agent-based modelling for social-science questions together with excellent communication skills are required. Only online applications received before midday on the 9th February 2025 can be considered.