Kempner Research Fellowship 2027: $100,000 Salary and Research Support for AI and Intelligence Scientists

Kempner Research Fellowship 2027: $100,000 Salary and Research Support for AI and Intelligence Scientists

The Kempner Institute at Harvard University is accepting applications for its next Research Fellowship cohort. The fellowship is designed for early-stage postdoctoral scientists whose work advances the fundamental understanding of intelligence, including computational models, cognitive theories, neurobiological approaches, artificial intelligence, machine learning, computational biology, and high-dimensional neural or behavioral data analysis.

For fellows beginning in fall 2027, the institute offers a salary of $100,000 per year plus benefits for up to three years. Fellows can also receive up to $10,000 per year for professional or research expenses. The position includes dedicated office and collaboration space, access to the Kempner Institute’s computing resources, and a community of scientists working across intelligence, AI, machine learning, and related fields.

The application deadline is October 1, 2026 at 6:00 p.m. Eastern time. Applicants should be in the final year of their Ph.D. at the earliest and able to formally complete, defend, and receive the degree by the fellowship start date. For the fall 2027 cohort, the Ph.D. must be received no earlier than September 1, 2024 and no later than September 1, 2027. The institute says successful candidates generally begin on September 1, 2027.

This is a research fellowship rather than a generic postdoctoral job listing. The application needs to show both an original research direction and a credible reason that the Kempner Institute is the right setting for it. The strongest candidates will make the connection between their proposed work, the institute’s mission, potential mentors, its computing environment, and its research community concrete.

Key details

What the fellowship provides

The financial package is unusually clear. Kempner research fellows receive a $100,000 annual salary plus benefits for up to three years. In addition, the institute provides up to $10,000 per year for professional or research expenses. The official page does not describe the additional support as an automatic unrestricted cash payment, so applicants should treat it as a research and professional-expense allowance and confirm the eligible uses during the appointment process.

The fellowship is intended to provide significant time and flexibility for research. Fellows have access to a state-of-the-art computing cluster and to engineers who support the institute’s machine learning infrastructure. They also receive dedicated office and collaboration space in the new Kempner Institute facility at Harvard’s Science and Engineering Complex.

The intellectual setting is part of the award. Fellows join a community of faculty, staff, students, and other early-stage scientists studying intelligence from different angles. The institute describes the fellowship as a place for collaboration, exploration, and intellectual freedom, but that freedom comes with participation expectations. Fellows are expected to focus full time on their research while also contributing to the Kempner community through activities such as mentorship, educational work, selection committees, or event organization.

Financial support is therefore only one part of the fit. Someone looking for a low-contact salary bridge may be disappointed if they do not want to participate in teaching, mentoring, community programming, or institute activities. Someone who wants a research environment with computational infrastructure and active interaction across fields may find the structure valuable.

Who should apply

The fellowship is aimed at early-stage postdoctoral scientists. The institute wants candidates with demonstrated ability and creativity whose work falls within its broad research focus. The page names two major groupings: foundations of intelligence, including mathematical and computational models, cognitive theories, and the neurobiological basis of intelligence; and engineering or scientific work on AI and machine learning, including new methods, models, LLMs, generative AI, computational biology, and analysis of high-dimensional neural and behavioral data.

The research is expected to be computational rather than bench-based. That does not mean every applicant must work on the same kind of model or use the same technical stack. It means the proposed fellowship research should fit a computational account of intelligence, AI, machine learning, or related data-driven scientific questions. A proposal centered on wet-lab experiments without a substantial computational research core is unlikely to match the stated program.

The degree timing is a hard screen. A candidate must be at least in the final year of Ph.D. training and must be able to complete, defend, and receive the Ph.D. by the fellowship start. For fall 2027 applicants, the degree must be received between September 1, 2024 and September 1, 2027. Applicants with a degree date outside that window should not rely on the general phrase “early-stage” to assume eligibility.

Candidates also need to propose two or three potential mentors from current Kempner faculty or from Harvard faculty working in machine learning, artificial intelligence, or computational neurobiology. An existing relationship is not required. The mentor list is still a substantive part of the application because it shows that the candidate has considered where the work could fit and how the institute’s expertise would shape the project.

Research fit and mentor selection

The best way to choose potential mentors is to start with the project rather than with prestige. Identify the technical or conceptual problem the fellowship would address, then find two or three faculty members whose work connects to different parts of that problem. One mentor might provide expertise in computational neuroscience, another in machine learning methodology, and another in theory or representation learning. The list should tell a coherent story about the proposed collaboration.

Applicants do not need to have secured a mentor before applying, but they should still read faculty research descriptions carefully. A proposed mentor who works in the same broad field is not automatically a strong fit. The application should explain what the candidate would learn from the institute and what perspective or method the candidate would bring to the community.

The proposal also needs to fit the Kempner mission without becoming a generic statement about AI. The institute is interested in the fundamental understanding of intelligence. A project that applies an existing model to a convenient dataset may be worth pursuing, but the application must make the underlying intelligence question visible. Explain whether the work tests a theory, develops a new computational method, studies a mechanism, or creates a model that could clarify how intelligent behavior is represented or learned.

Application materials

The required materials are a CV, a list of two or three proposed mentors, a research proposal of no more than three pages or 1,500 words excluding references, a personal statement of no more than two pages or 1,000 words, and three references.

The CV should highlight publications and conference abstracts relevant to the Kempner mission. Do not assume that a long publication list will communicate the research story by itself. Make the contribution of the most relevant work easy to identify, especially if the applicant’s background spans neuroscience, computer science, mathematics, biology, psychology, or another field.

The research proposal should outline the research to be performed at the Kempner Institute. Within three pages, it needs to establish the problem, the proposed approach, what is new, and why the work is feasible during a fellowship of up to three years. A proposal can be ambitious, but it should not depend on an undefined collaboration or a computing resource that the application never describes.

The personal statement must cover the applicant’s Ph.D. research and past accomplishments, how the work fits at the interface of Kempner research areas, and how the applicant will interact with the Kempner community. This is the place to explain the fellowship’s professional logic. Show how the institute’s people, infrastructure, courses, workshops, and collaborative environment would change the applicant’s next research stage.

Three referees are required. They should be able to describe the applicant’s work and potential for future discoveries. The application is not complete until all letters have been received, although the institute says references have one week after the application deadline to submit their letters. Applicants should still request letters well before the deadline and monitor the portal rather than assuming a referee’s promise means the letter has arrived.

Selection and interview process

Kempner says fellows are selected based on four main considerations: alignment of the proposed research with the institute’s mission, vision, and values; the extent to which the proposed research is novel; the individual’s potential to add value to and benefit from the Kempner community; and a record of successfully pursuing ambitious projects independently.

Those criteria reward more than technical competence. A candidate may have excellent methods but still need to explain why the project belongs at Kempner and how it differs from the candidate’s previous work. The phrase “add value to and gain benefit from the community” is also reciprocal. Applicants should identify the conversations, collaborations, mentoring, or educational contributions they would pursue rather than treating the institute as a computing facility with no social dimension.

Candidates will be invited in mid-November to present both past and proposed work. Presentations and interviews are virtual. The published dates are November 12, November 13, November 19, and November 20, 2026, with different time windows on those days. Applicants should keep those dates available and be prepared to explain the proposed research clearly to a mixed institute audience.

How to prepare before October 1

Begin by writing a short project brief that answers five questions: What is the intelligence problem? What is known already? What would be new? What evidence or model would test the idea? Why is Kempner the right environment? If the fifth answer is interchangeable with any university, the project fit needs more work.

Next, build a three-year research arc, even if the first proposal focuses on one tractable study. The first stage might establish the model or dataset, the second might test the central claim, and the third might extend, validate, or connect the work to a broader theory. The plan should be specific enough to establish feasibility without pretending that every result is already known.

Then select potential mentors and read their recent work. Note the exact connection for each person: a method, theory, dataset, experimental collaboration, or community activity. Those notes can become part of the proposal and personal statement, but they should not be copied as generic name-checking.

Finally, ask three different readers to review the materials. A subject expert can test the science. A researcher from a neighboring field can test whether the proposal is understandable. A colleague who knows the applicant can test whether the personal statement shows independence, judgment, and a genuine reason for joining Kempner. With strict page limits, clarity is part of the evidence.

Common mistakes to avoid

The first mistake is treating the fellowship as a general AI postdoc. Kempner is looking for work related to the fundamental understanding of intelligence, not every project that uses machine learning. Make the scientific question visible and explain the relationship to the institute’s mission.

The second is proposing a bench-based project without a meaningful computational center. The official eligibility language says the research focus is expected to be computational rather than bench-based. If the project combines biological experiments and modeling, explain the computational work precisely and make clear which part the fellow would lead.

Another mistake is naming mentors only because they are prominent. The application does not require an existing relationship, but the candidate still needs a plausible reason to list each person. A thoughtful list of two or three complementary potential mentors is more useful than a long list of loosely related names.

Applicants should also protect the proposal and personal-statement limits. Three pages or 1,500 words is not enough for a full dissertation plan, and two pages or 1,000 words is not enough for a complete autobiography. Select the evidence that supports the fellowship case. Use the CV for the full record, the proposal for the future research, and the personal statement for the connection between past work, Kempner fit, and community participation.

Finally, do not leave references until the last day. The application is incomplete until all three letters arrive, even though referees have one additional week after the deadline. Confirm each referee’s email address, send a focused briefing, and check the application system before the October 1 cutoff.

Frequently asked questions

Is the $100,000 support a stipend or a salary?

The official page describes it as a $100,000 annual salary plus benefits. The fellowship also includes up to $10,000 per year for professional or research expenses.

How long is the fellowship?

Fellows serve for up to three years. Successful candidates generally begin on September 1, 2027.

Can someone who already has a Ph.D. apply?

Yes, if the degree date fits the published early-career window. For the fall 2027 cohort, the Ph.D. must be received no earlier than September 1, 2024 and no later than September 1, 2027. Candidates must also fit the institute’s research and computational-focus requirements.

Do I need to have a mentor’s agreement before applying?

No. Applicants are asked to provide two or three proposed mentors, and the institute says candidates do not need to have established a relationship with those faculty members in advance. The proposed mentor list should still be based on real research fit.

What happens after the application is submitted?

Candidates will be invited in mid-November to present past and proposed work and then interview with members of the Kempner community. The presentations and interviews are virtual, with published dates on November 12, 13, 19, and 20, 2026.

Is the fellowship only for computer scientists?

No. The stated research areas include mathematical and computational models, cognitive theories, the neurobiological basis of intelligence, AI and machine learning methods, computational biology, and analysis of neural and behavioral data. The common requirement is a computational research focus connected to intelligence or related institute priorities.

Official application link and next step

Start with the official Kempner Research Fellowship page and use its current Apply now link for submission. Before writing, check the institute’s mission, values, faculty research areas, and open-science terms. The page says fellows who receive financial support must sign an Open Science Participation Agreement covering research integrity, data access, intellectual property, software, inventions, reagent sharing, publication, and collaboration.

The immediate preparation checklist is short but demanding: confirm the Ph.D. timing, choose two or three plausible mentors, outline a novel computational project, draft the three-page proposal and two-page personal statement, and secure three referees. Submit by October 1, 2026 at 6:00 p.m. Eastern time. A strong application will show an independent research record, a focused next project, and a clear two-way fit with the Kempner community.