Cambridge Residency Programme – AI Efficiency

Key Skills

Aartificial intelligence, Design, Large Language Models - LLMs, Machine learning techniques, NeurIPS, Python Programming, PyTorch

Job Description

We are looking for a Postdoctoral researcher to join us as part of a two-year AI Residency Program in Cambridge, UK. This is a close collaboration between M365 Research and Microsoft Research to advance the state of the art in AI efficiency. 

Generative AI is revolutionizing how we create, share, work with, and consume content. At Microsoft, we run the biggest platform for collaboration and productivity in the world with hundreds of millions of consumer/enterprise users. 

Tackling AI efficiency challenges is crucial for delivering these experiences at scale.  

Within our Microsoft wide Systems Innovation initiative, we are working to advance efficiency across AI systems, where we look at novel designs and optimizations across AI stacks: models, AI frameworks, cloud infrastructure, and hardware. We believe that enabling cross-layer optimization is the key to achieve a step-function improvement in AI efficiency and to provide generative AI experiences to more users globally. 

In this Research Residency you will work with researchers and engineers to develop new methods that can be deployed up and down the AI stack to deliver efficiency at scale. You will conduct research into one of the following themes: 

LLM system efficiency including but not limited to: efficient in-context learning, hybrid inference and efficient in-domain finetuning, to improve inference efficiency of LLM systems while also enhancing their accuracy and generalization capabilities.  

Model level efficiency. By exploiting mathematical structures that are common in large AI systems, you will develop new post-training compression, optimization and quantization techniques that will drive efficiencies at the model level. This may include for example cache compression to overcome memory limitations in LLM deployments. 

We are looking for enthusiastic researchers to help advance the state of the art of Efficient AI. The ideal candidate will have background in machine learning and natural language processing, with experience in model distillation/compression, adaptive computation, architecture optimisation, and/or approximate inference, and the motivation and ambition to apply this in a real-world setting.

Qualifications

Required/Minimum Qualifications:  

• PhD in computer science, math, physics, or related field
• Capability for rapid and accurate prototyping in Python.
• Experience with ML frameworks such as pytorch.

Preferred/Additional Qualifications:

• Knowledge of state-of-the-art Large Language Models (LLMs) 
• Publishing track record in area of efficient machine learning, e.g. NeurIPS, ICML, ICLR.

Responsibilities

• Conduct novel research to advance the state-of-the-art in inferencing and training optimization across the stack for Large Language Model / Generative AI model systems.  
• Work with a small group of fellow research scientists and product engineering teams to execute practical solutions for real-world impact.  
• Publish and contribute to top scientific conferences and journals. 

Benefits/perks listed below may vary depending on the nature of your employment with Microsoft and the country where you work.

• Industry leading healthcare
• Educational resources
• Discounts on products and services
• Savings and investments
• Maternity and paternity leave
• Generous time away
• Giving programs
• Opportunities to network and connect