Postdoctoral Researcher (Foundational AI: Theory and Architectures)

The Machine Intelligence team at Microsoft Research Cambridge conducts foundational machine learning research across learning algorithms, model architectures, reasoning systems, and efficient AI. 

We are seeking a highly collaborative Postdoctoral Researcher with expertise in one or more core areas of AI/ML, strong communication skills, and a commitment to impactful fundamental research. We are interested in fundamental questions underlying modern AI systems, such as: 

• Principles governing learning, representation, reasoning, adaptation, and generalization;
• The interaction between model architecture, learning algorithms, data, and computation;
• Foundations of efficient, scalable, and capable AI systems;
• Mathematical understanding of phenomena observed in modern AI/ML systems;
• New algorithmic, theoretical, or architectural approaches that advance AI.  

• Contract Type: Resident (2 Years) 
• Location: Cambridge, UK

• Develop an original research program that advances fundamental understanding or methodology in modern artificial intelligence. 
• Formulate and investigate important research questions using theoretical analysis, algorithm or architecture design, empirical study, or an appropriate combination of approaches. Derive, design, implement, validate and iterate on new architectural through controlled and large-scale scaling experiments.
• Collaborate with researchers working on model efficiency, reasoning systems, learning algorithms, and large-scale experimentation.  
• Pursue an independent research program and contribute to the team's broader scientific direction.  
• Disseminate research through peer-reviewed publications, conference presentations, open scientific engagement, and collaboration with the broader research community. 

Required/Minimum Qualifications:  

• A PhD (completed or near completion) in Machine Learning, Computer Science, Mathematics, Statistics, or a related field.
• Expertise in one or more sub-fields of AI/ML, evidenced by top-tier publications and/or experience. 
• A strong record of original, peer-reviewed research published at leading venues in AI, machine learning, or theory,  such as NeurIPS, ICML, ICLR, COLT, STOC, FOCS, or SODA. 

 Preferred/Additional Qualifications: 

• Expertise in one or more areas relevant to theoretical and/or practical aspects of  modern foundation models, including but not limited to: optimization and training dynamics, LLM architecture, representation learning, attention mechanisms, test-time computation or training, reasoning, memory, adaptation, agentic interaction, efficiency, scaling or machine learning theory.  

• Demonstrated ability to formulate, investigate, and iterate challenging research questions independently. 

• Excellent written and verbal communication skills, with the ability to explain complex ideas to a broad scientific audience. 

• Hands-on experience designing, implementing, training, and evaluating AI/ML models using a modern framework such as PyTorch or JAX. 

We value depth, originality, and transferable skills in a candidate's chosen area. We do not expect applicants to cover all the topics or methods listed above, and we welcome candidates whose primary contributions are theoretical, empirical, or methodological.