Eric and Wendy Schmidt AI in Science Global Faculty Fellowships

Eligibility criteria: 

• The institutional target (except for USP) is that one fellow per year on average will be selected. 
• Candidates can be from any department – this call is for candidates from any department within remit that seeks to change their science/engineering using AI (it is not only a fellowship for those from Computing – the large majority of fellows are NOT from computing departments and do not have degrees in computer science). 
• In order to be eligible to apply, applicants must be early/mid-career and be within the AI in Science remit.
• Under the AI in Science Remit, ‘AI’ is interpreted very broadly, that is including topics in Bayesian Inference and Robotics, with ‘Science’ covering any typical topic in Natural Science and Engineering.
• These fellowships are not suitable for research into generic AI with general application - candidates must be aiming to substantially advance a very particular area of science. Applicants could view themselves as AI researchers tackling a particular piece of Science or Science researchers using AI to transform their area.
• A deep knowledge of AI is not a precondition for this fellowship: only an appreciation of the need for AI and a willingness for skill acquisition in AI. Candidates do not need to come from Computing departments. 
• Full details of remit can be found in the section below. There is no flexibility about remit.

To apply please send your CV including publications to your home institution. 

Additionally, please submit a single file with: 

• Publication Elaboration: a 1 page, or less, note outlining the contribution of up to three papers by the applicant. This should be suitable for a general scientific reader.
• Research Proposal Summary: a 1 page or less, summary of the proposed research suitable for a general scientific audience including the title of the research project. Particular attention will be paid to this summary. It should answer the question of why/how this application of AI will be transformative for the target area of science. The proposal should start by mentioning the applicant’s proposed department and at least one faculty member at Imperial who would support the visit and act as a host. Hosts must be contacted in advance of the application and will need to supply a letter of support form. It is not essential that the host be a very close fit to the proposed research, entirely independent research efforts are welcomed, but a collaborative relationship is likely to make the science more credible and help with integration.
• Research Proposal: 3 pages or less proposal that explains why and how the proposed research could be transformative for a particular area of science. It can be structured around background, a small number of hypotheses/aims, and work packages. It can be assumed that the reader will first read the Summary and so content need not be repeated.
• Training Plan: a ½ page or less plan, identifying any particular skills that need to be acquired for the proposed research to succeed. Training is a key part of the proposed fellowship, whether helping an AI expert master a scientific topic or a scientific topic expert advance their AI skills. Deep expertise in AI (or the particular Science area) is not a pre-requisite: the minimum level of AI/Science experience is that needed to credibly articulate a plan for how AI will advance Science.
• Fit to AI in Science Remit: a ¼ page or less outline of how your proposal fits within the AI in Science remit.
• Kindness Statement: a ¼ page or less outline of your view on the need for kindness among scientists.
• A letter of support form from a proposed academic host at Imperial College London (see below)
• A letter of support form from Home Head of Department (see below)

How will Fellows be selected?

• Each partner institution will formulate an internal review committee that reviews and ranks the proposals submitted by candidates from their institution.
• The committee will then submit a ranked list of the top three (or less) candidates (Note: USP will submit the top six candidates) to the selection committee at Imperial.
• The selection committee at Imperial will conduct interviews as needed and would seek to select on average one successful candidate each from African Institute for Mathematical Sciences, Indian Institute of Science, National Centre for Biological Sciences, India, University of Ghana, Universiti Teknologi Malaysia and on average three successful candidates from USP.
• Applicants must have identified and contacted researchers at Imperial in relevant research fields and secured interest from them in the form of a letter of support.

Applications must be supported by a host who is a permanent member of staff within the Academic and Research Family at Imperial.

An Imperial host MUST be secured prior to submission. Applications that are not fully supported by an identified Imperial host will not be considered. 

Useful links below (including the Imperial AI Network and the I-X list of academics) within the College’s website can help you find a potential host.   

• View Imperial AI Network  
• View I-X Staff Profiles 
• View a list of academic departments by Faculty 
• View an A to Z of Departments and Centres 
• View the Imperial College staff directory website 
• View the Research and Innovation webpage 
• View the College news webpage 

Please note: hosts are not obliged to accept a request for support. If they choose not to support a potential applicant, it is at the host's discretion if they choose to give feedback. 

Here ‘AI’ is interpreted very broadly, e.g.: including topics in Bayesian Inference and Robotics; ‘Science’ covers any typical topic in Natural Science and Engineering (Epidemiology, Biology and basic science in biomedicine are included but clinical medical themes are not covered, including conventional medical imaging). Example topics include Bayesian optimization for molecular or materials design; machine learning for single cell data; physics-based ML for turbine design; satellite imaging to predict disease risk; and Astrostatistics. These posts are not suitable for research into generic AI with general application: candidates must be aiming to substantially advance a particular area of science. Applicants could view themselves as AI researchers tackling a particular piece of Science or Science researchers using AI to transform their area. A deep knowledge of AI is not a precondition for this fellowship: only an appreciation of the need for AI and a willingness for skill acquisition in AI. 

We do include epidemiology but outside this no: the grant’s principal orientation is towards the natural sciences and engineering. 

‘Science’ covers any typical topic in Natural Science and Engineering (Epidemiology, Biology and basic science in biomedicine are included but clinical medical themes are not covered, including conventional medical imaging): fellows are expected to have their principal host in the Faculty of Natural Sciences and Faculty of Engineering. We are pleased to support medical research that attempts to characterise a biological system/process including a pathophysiological process/state. Our centre does not seek to support translational medical research or the development of therapeutic technologies. A proposal can certainly connect to therapy and technology, but this should not be the principal objective of the fellowship.

Yes! Deep knowledge of AI is not a precondition for applying -- the fellowship can help you pivot into acquiring AI skills to advance your disciplinary area. You might want to pick your host and design your training program accordingly. 

Yes! Deep knowledge of the Scientific area is not a precondition for applying -- the fellowship can help you pivot into acquiring deeper subject area expertise. Please pick only one area to target. You might want to pick your host and design your training program accordingly.

Yes! Critically you should view this fellowship as an opportunity to pick one application area and pivot into gaining a much deeper mastery of it -- you cannot cover multiple application areas in different fields within the fellowship. You might want to pick your host and design your training program accordingly. 

Yes! You would need to cover how the AI you do is somewhat distinctive and not very widespread in the field. 

In some scientific subfields there are now very well-established AI tools, and the use of these particular AI tools could be considered ‘completely routine,’ ‘standard practice appearing in almost all papers within the scientific subfield’ or extremely widely used’: these fellowships are broadly intended to unlock new areas of science through AI. While we appreciate that this is a grey area, we ask candidates to clarify if they are proposing a different set of AI tools to those that are very well established (or developing new AI tools) with the goal of unlocking new science; or taking some kind of distinctive approach; or that their approach is not very standard practice. A particular use of AI that is standard practice in one subfield might be rare in another (perhaps close) subfield: for any given AI approach our fellowship favours the application of an AI approach to a subfield in which it is more rarely used.  

This fellowship is not about supporting researchers that already do generic AI to continue doing more of their work without strong engagement with specific disciplinary concerns outside their current area in generic AI. AI includes topics like inference, robotics, control, signal processing, AI hardware etc. Science is as above and explicitly excludes topics within ‘AI’ as defined. A project is not AI in Science if e.g., it uses new ideas from signal processing to develop robotics or new Machine Learning to improve robotic control or Bayesian Optimization to Improve AI Hardware -- these would be ‘AI in AI’ projects. Research projects that e.g., seek to develop better AI by studying the brain are outside remit (Science for AI). 

An answer through examples is easier. A researcher in algebraic geometry might want to use the fellowship to pivot into acquiring AI skills to allow development of theorem proving assistants in algebraic geometry. Or a fellow might use large libraries of solutions to equations and develop AI tools to make sense of them. In both cases insight into specific mathematical structures is gained by using AI and thus is within remit. A possible heuristic is to ask -- “am I trying to use AI to (ultimately) make a mathematical discovery”? A researcher with a focus on PDEs might use the fellowship to use AI (e.g., physics-based ML) for a specific scientific or engineering challenge (nominal engagement with a specific area is insufficient). Many areas like Astrostatistics, Biomathematics, Climate Mathematics or Engineering Mathematics often already straddle both AI and Science and so will often be straightforwardly within remit (providing that AI is used to advance some particular area of science). Statistics with generic application areas is not in remit; method development with generic application areas is not in remit.