Program Scope
The definition of AI for this program is broad: it includes machine learning, robotics, Bayesian inference, and simulation. However, this does not mean that everything in these topics is AI. For example, servo-motor design may be important for robotics, but is not AI; and numerical methods for finite element analysis can be useful in many scenarios, but are not AI. On the other hand, our definition of AI is not limited to these four terms: it also includes many additional areas of expertise important for advancing AI in science and engineering that do not fit within these four buckets. For example, data integration can be an essential precursor to the use of AI methods, and hence would be in scope. Similarly, using AI to automate research workflows is a topic of interest.
Physical sciences, mathematical sciences, earth and environmental sciences, and engineering, as well as biological and biomedical science research that uses AI methods to understand the basic biological processes, disease mechanisms and propagation (in silico, in vitro and animal models) is within scope. However, biomedical or healthcare research with a focus on developing or improving disease preventions, cures or treatments is out of scope.
A note about computer science. Much of AI is in computer science, and the participation of AI methodologists is essential for the success of this program. However, computer science is not a science or engineering domain within the scope of this program.
Note: MIDAS has been, and remains, a strong supporter of data science and AI in other disciplines, including biomedical, translational and healthcare research, and social sciences. We provide postdoctoral training in these disciplines through our sister program, the Michigan Data Science Fellows (DS Fellows) Program, which focuses on responsible data science and AI to improve their scientific and societal impact, and data science applications in social science and healthcare research. Potential applicants whose research is out of scope of the Schmidt AI in Science program may consider applying to the DS Fellows program
Program Structure
Appointments will begin between August 1 through September 15, 2025. Appointments will be for a one-year (12 months) initial duration, and renewable for one more year. Most appointments are expected to have a total duration of two years. The renewal for year 2 will be based on satisfactory performance in year 1. Extension of the Fellowship beyond two years will be competitive and granted only in exceptional circumstances.
No. Fellows must work in-person on the University of Michigan campus. A hybrid work schedule may be possible in rare cases, based on mutual agreement between the Fellow, their mentors, and the program directors.
The program will pay a competitive salary ($81,000 annually for 2025-26) plus benefits. Travel to annual Schmidt Sciences events will also be covered. In addition, there may be minor supplemental funding available for Fellows to attend AI training workshops / short courses and professional conferences.
The mentors cover all costs of the Fellows’ research, including research space, lab facilities, computers, computing resources, data access, travel to research conferences (at least two each year), publication fees, and other necessary costs, such as visa fees when applicable.* Office desk space will be provided in MIDAS, but the faculty mentor may wish to make desk space available in their research group as well. The Science mentor (see below), by default, will be responsible to ensure the availability of all of the above; but the specific arrangements and shared responsibilities will be determined by the two mentors of each Fellow.
*For questions or concerns regarding the financial support of visa fees, mentors should contact the Program Team, schmidt-aim@umich.edu.
A faculty mentor can be any faculty member on the tenure track, clinical track, or research track with a full-time appointment at the University of Michigan in any of the Ann Arbor, Flint, or Dearborn campuses. They should be a MIDAS affiliate or have applied to be one.
For the sake of the program’s scientific diversity, each faculty mentor may not have more than one Fellow in the program at any given time. In other words, if a mentor of a current Fellow would like to be the mentor of an applicant, the current Fellow’s end date should precede the applicant’s start date (if accepted).
Faculty members at other universities may not be mentors. However, they may be members of the Fellowship Committee if they are able to fulfill the obligations of the committee.
Each Fellow is expected to have two mentors, a Science mentor and an AI mentor. The mentors will play an important role in helping the Fellows become outstanding researchers. They should guide the Fellows to develop research plans, identify challenges, and ensure that the Fellows develop adequate domain knowledge and AI skills. The mentors should also support the Fellow’s participation in the postdoc program activities and be active members themselves of our postdoc and mentor community. How the two mentors jointly fulfill their duties, for example, whether to designate a primary mentor, how exactly to divide responsibilities, and who covers which portion of the research costs, will be decided between the mentors. MIDAS will require letters of support from both mentors to expressly commit to their responsibilities. Additional mentors are allowed but not required
What we really look for is strong mentoring for both the research direction (and the specific project) and AI skills training. Oftentimes one mentor has stronger AI expertise, and the other has stronger domain expertise. In such cases it’s easy to designate the two mentors’ roles. However, sometimes both mentors are strong in AI and in domain science. In these cases we let the mentors make the designations with the understanding that, with their combined expertise, they will provide sufficient mentoring in both domain research and AI training to the Fellow. In the application, this should be clearly explained; Fellow and mentor applicant teams will be reviewed more favorably if the plan for AI and science domain mentoring is clearly explained and justified.
After a Fellow starts in the program, they and their mentors, program directors, and one additional faculty member (preferably a close collaborator with the Fellow) will form the Fellowship Committee. The committee will meet every six months to review progress in research and training, and provide guidance to the Fellow. The Fellow can revise their Individualized Development Plan based on the committee’s recommendations.
Diversity, equity and inclusion (DEI) is a core value at U-M, at MIDAS, and in this postdoc program. The program values diversity in both the Fellows and the faculty mentors. The program seeks demographic diversity across race, gender, ethnicity, personal background and beliefs; professional diversity such as the PhD granting institutions of the Fellows and their career directions. The program also seeks to support Fellows working in a diversity of fields in science and engineering on a wide range of research questions.
The program will be advertised through multiple channels to reach diverse applicants. During the application process, the program team will help applicants connect and communicate with potential faculty mentors, to ensure equity among applicants from all backgrounds. Reviewers of applications will be required to follow anti-bias best practices developed by the university. The Fellowship selection committee members are expected to have gone through anti-bias training. In addition, all candidates are required to submit a DEI statement. The program will also develop a rubric to help faculty mentors mitigate unconscious bias when they make subjective judgments.
After the Fellows join the program, we will effectively use the Individualized Development Plan and the Fellowship Committee to ensure equity in their training and career preparation. We will promote inclusion strongly in collaborative learning and community building activities.
A key aspect of the program is skill development for the Fellows. The program has many components to facilitate the Fellows’ skill development. However, such skill development will be effective only if Fellows are motivated. Passive reliance on program support is unlikely to suffice. As such, we explicitly require an Individualized Development Plan for each Fellow, which will be crafted and revised by the Fellows, their mentors and their Fellowship Committees. We will also track the progress of the Fellows closely.
The program has the following built-in activities.
Onboarding bootcamp. The program organizes a weeklong boot camp for each new cohort of Fellows, along with continuing Fellows and mentors (mentors may attend only a portion). The boot camp will consist of technical tutorials, career guidance sessions, and research presentations.
“AI carpentries”. The Fellows across all of our training programs form “AI carpentries” – small groups based on common research interests and methodologies. Members of each carpentry will be required to work together at MIDAS collaboration space one day a week. Faculty mentors will take turns attending and guiding the work.
Weekly research meetings. All Fellows attend these meetings to discuss projects, pitch new ideas, share interesting papers, practice their presentations, and carry out other group learning activities. Other postdocs and faculty members on campus will also be invited to these meetings.
Expectations
First and foremost, the Fellows should be committed to successfully implementing their research plan and the Individual Development plan. But equally importantly, they should be committed to being active members of the postdoc program and the U-M AI in Science research community.
Each Fellow will have a desk in MIDAS’s postdoctoral space, and is expected to spend at least 50% time in this space. Each Fellow should also spend an appropriate amount of time with their faculty mentors’ research groups (based on agreements between the Fellows and mentors).
After onboarding, each Fellow is required to set up a Fellowship Committee as described above. Committees will meet twice per year and will begin with a presentation of progress made by the Fellow.
All Fellows are required to actively participate in the collaborative learning activities.
All Fellows are also expected to participate in programmatic activities that can help build their professional service experience. They may choose to participate in training activities as instructors; or organize research connection events; or develop algorithms, tools, research protocols that can be shared with the campus research community. These activities should account for, on average, 10% of the Fellows’ effort.
During the fellowship, if the Fellow proposes a significant change to their research or their in-residence status, the Fellow must submit a request to program leadership with a revised Individual Development Plan (IDP), describing how the proposed change supports their overall research plan and career preparation. Examples could include a period of a few months in which the fellowship is split with specific duties on a mentor’s new research grant, or a short-term leave for the Fellow to take advantage of a research opportunity overseas. It is important that such requests come to program leadership from the Fellow, not from the mentors.
Schmidt Sciences, the funder of the postdoc program, will organize events across the postdoc programs that it funds at multiple universities. Fellows should abide by their attendance requirements.
After they leave the program, Fellows are expected to provide their CV to program management at MIDAS annually for at least five years, as well as an annual summary of their research activities and accomplishments, and how they advance the application of AI to enable research breakthroughs.
Mentors are expected to devote sufficient time to provide research and career guidance to the Fellows, meet with them regularly, attend Fellowship Committee meetings every six months, and help Fellows seek opportunities for career advancement. They should also support the Fellows’ participation in activities of the postdoc program and their professional service. They are also required to cover all costs of the Fellows’ research, other than Fellow salary and benefits, as discussed above in “What do the mentors pay for?” and “What is the role of faculty mentors?”.
In addition, mentors are expected to actively participate in program activities. A rough estimate of time commitment for some of the events is as follows:
- Any attendance requirement set up by the funder (for example, they may invite some faculty mentors to their annual conference).
- Annual bootcamp: each mentor is expected to attend parts of this weeklong event.
- Weekly research meetings: occasional attendance.
- Fellow selection: Several hours of application review once a year.
- Being a champion for the program for Fellow recruitment and for the program to have a broader impact on the adoption of AI for science and engineering research on campus (a few hours to a few days a year).
Qualifications and the Application Process
The candidate must have a PhD degree prior to the start date of the Fellow appointment, but within five (5) years. The PhD can be in any field. (Note the science and engineering domains listed above that are in scope: these apply to the area of proposed work, but not to the field of PhD degree). It is acceptable for the candidate to already have a postdoc position, whether at U-M or elsewhere (but please see below about U-M candidates). However, the candidate should not have had a faculty appointment anywhere that is equivalent to a North American university’s tenure-track position.
The ideal candidate will have training in both AI (at least one sub-area) and the domain discipline. However, this may not be the case for many candidates. We require that a candidate have deep expertise in at least one of the two: either AI methodologies or the domain discipline, in conjunction with demonstrated interest in the other. The key requirement is evidence of excellence and a passion for the proposed research direction at the intersection of two disciplines.
See the program website for information about how and when to apply. For the 2025 cohort, the application will open on September 19, 2024 and the priority deadline is November 23, 2024. Fellowship offer distribution is expected to begin in mid-February 2025.
To develop a true Fellows cohort each year, and to allow synchronized programmatic activities and skill building for each cohort, the program has a target start date window of August 1 to September 15. Some exceptions may be possible, in special circumstances, for example, due to visa processing. .
Yes. But you will need to complete all degree requirements before you can be appointed as a postdoctoral fellow and start on our expected date.
If you have already successfully defended your thesis, but your degree has not been officially conferred, you are eligible to apply as long as the conferral will happen before or shortly after your start date. You will need an official letter from the appropriate University authority stating that you have completed all requirements and your degree will be conferred on such and such date.
Yes, the program is open to international applicants. There is no citizenship requirement for the applicants. Accepted Fellows will need to obtain the appropriate visa before joining the program. If for any reason a visa is not granted, the acceptance will be voided.
It depends on when you submit the application. Applications that propose a direct continuation of one’s previous or current research, or a continuation of previous or current mentorship, as of their application submission date, would be disqualified.
It depends on when you submit the application. Applications that propose a direct continuation of one’s previous or current research, or a continuation of previous or current mentorship, as of their application submission date, would be disqualified.
The candidate should seek at least one faculty mentor either in AI methodology or in an application domain. If one faculty member agrees to be a mentor, they may work with the candidate to find a second mentor. Candidates may find it useful to use the MIDAS website to search through the 580+ affiliated faculty, although faculty mentors need not be selected from this list. Here is a list of faculty members who have expressed interest in being AI mentors.
However, candidates should feel free to reach out to faculty members not on this list.
If needed, the program team will also help candidates identify mentors.
Yes. Just like mentors can help their trainees prepare an application for an NIH or NSF funded research fellowship, or a fellowship funded through a foundation, mentors can and are encouraged to help a candidate prepare the application to our Fellowship program. Specifically, we encourage mentors to discuss extensively with the candidate both the proposed research project and the Individualized Development Plan.
Yes. We will not dismiss or downgrade an application on account of a mentor who expresses support for multiple fellow applicants. So it is OK to support multiple applicants.
However, keep in mind that we seek diversity along many dimensions. Having two Fellows working with the same mentor goes against that, and so we will not want that in our final portfolio. So, among multiple candidates who apply with the same mentor, we will accept at most one. For the sake of the program’s scientific diversity, faculty mentors of the current Fellows can be mentors of new applicants only if the end date of their current Fellow precedes the start date of the new applicant.
Our advice to prospective mentors: if you think one applicant is clearly stronger than another, go only with the stronger one: you are unlikely to get a second. However, if you cannot decide between two or more applicants, please feel free to support all and let us evaluate.
The nature of research is that there are many unknowns and it is okay if the proposed research project evolves during the Fellowship. However, the research statement is one of the most important selection criteria. The research statement should include a compelling vision and a plausible approach to achieving that vision. Note that the program seeks to promote research that is transformative and creative, with the potential to spur significant advances in the application of AI in science domains.
No need to be comprehensive. The IDP in the Fellowship application should demonstrate that the applicant and their mentors are aware of the applicant’s training needs and are committed to meeting such needs. We also do not favor one type of training over another. For example, we will not require Fellows to take formal classes. The IDP should be effective and feasible for each Fellow. The key evaluation criterion here is whether a thoughtful approach has been taken to consider the Fellow’s background, the skills that will be needed to complete the proposed work and to round out the Fellow’s training to apply AI methods in a science domain, and the Fellow’s career plan.
After a Fellow starts in the program, the IDP will be used as a guide and will be referred to in Fellowship Committee meetings. Fellows will work with the mentors and the Fellowship Committee to revise the plan based on their training needs and progress.
Responsible AI includes not only scientists’ responsibility to society, but also their responsibilities to science. From the rigor of study design and the explainability of models, to the unbiased use of data and algorithms, and to data sharing and making a research project fully reproducible, responsible AI is critical to all applicants, because all researchers need to take into consideration the effect of flawed data, brittle analyses, and irreproducible results.
We look for sincere and feasible ideas that contribute, even in small ways, to the diversity, equity and inclusion of the postdoc program, and to the AI in Science and Engineering research community on campus. A candidate’s life experiences can be described where they are relevant to DEI and a candidate’s past activities can be used to illustrate their contributions to promoting DEI.
Yes. These letters are similar to letters of support in training grants, and should be included in the application package. These are different from reference letters, which are usually confidential.
If you can secure three reference letters that can provide valuable information to the reviewers, you should. Having strong letters that can speak to your strengths and achievements will help you. However, we only require two letters. This can work well if one of your proposed mentors already knows you well, and can speak about both your future and your past in the letter of support.
Other Questions?
For questions, please contact schmidt-aim@umich.edu.