SoTS Research Residency

Real Problems, Real Stakeholders

Every team works on a real problem an external partner needs solved. Partners are typically researchers in the natural sciences: a university group, a government lab, a company, or an individual working scientist bringing AI into their research. Sometimes the problem is specific; often it is open-ended: where could AI help here, and how? The outcome is unknown in advance, to the team and the Residency Mentor alike.

There is no spec to hand off and build. A team has to understand both the science and the problem, map the existing research workflow, determine where AI genuinely helps and where it does not, then design, implement, and test a solution.

The Residency is built on a research-education methodology that Dr. Sergey Samsonau developed at NYU and refined over seven consecutive semesters, with more than 100 students completing 20+ team projects for external researchers at exactly this intersection. The methodology is published (arXiv:2210.08966). The Residency brings it to experienced high-school researchers, online.

• 7semesters refining the methodology
• 100+undergrad and grad students in the original program
• 20+team projects for external researchers

How It Works

• Teams of 5 to 6 Residents work with one partner across two consecutive semesters.
• Defined roles within each team: every Resident owns a distinct part of the work, such as AI and computation, domain science, data and experiment, literature and theory, or coordination and communication. Roles divide responsibility, not effort: everyone researches, everyone presents.
• Live online sessions with a professional Residency Mentor: once a week in fall and spring, twice a week in summer, 1 to 1.5 hours each.
• Daily independent work between sessions: roughly 10 hours a week in fall and spring, 20 hours a week in summer.
• Weekly presentations inside the cohort: defend your reasoning to sharp peers.
• Final deliverable to the partner: a result they can actually use, such as an improved research workflow, an application working scientists can run, or a novel AI perspective on the processes inside a lab.

A professional AI engineer guides Residents through this: how to bring modern AI and data science into an unfamiliar field, how to write research software that holds up, and what good practice looks like in version control, reproducibility, rigorous evaluation, and honest reporting.

Start Any Semester. Stay for Two.

The Residency admits new teams three times a year, fall, spring, and summer, on the same calendar as Research Labs. Every team commits to two consecutive semesters. The first semester takes a team through scoping, the literature, first results, and the first real obstacle. The second semester is where the team works through it. That is where the learning happens.

The first cohort begins July 1, 2026. Apply for the summer start, or for a later fall or spring cohort.

At the semester boundary, each team presents its progress to the partner in a midpoint review. Continuation into the second semester is by demonstrated work, the same standard as everywhere at SoTS.

Because cohorts overlap, the Residency always contains teams in their second semester working alongside teams in their first. Newer teams watch veteran teams present; veteran teams answer newer teams' questions. This is how a real research group works, and it is by design.

Who Gets In

Admission has two requirements, and both must be demonstrated, not claimed.

At least one year of original, self-driven research. A sustained investigation you formulated and carried yourself, in a SoTS lab or anywhere else. Polished or not. Successful or not. What matters is the trace of actual work: what question you pursued, how far past the first obstacle you chased it, what you did when the easy path closed. Where the work was done does not matter. What school you attend does not matter. Whether it won anything does not matter.

The ability to reason from first principles in at least one discipline. Physics, biology, chemistry, mathematics, computer science, or the organization of collective work.

The two requirements test different things. A year of work shows you can sustain a project. First-principles reasoning shows you understand a field deeply enough to apply it. A partner team needs both.

Admission is also a matter of composition. Teams need complementary strengths, so we admit the way a research group hires: qualified candidates first, then fit with the roles a team needs. Your prior research does not have to involve AI. A team needs domain scientists and rigorous experimentalists as much as it needs coders. A strong applicant may be offered a later cohort because this one already has the AI specialists it needs and no one who knows the science. That is a team-composition decision, not a judgment of the work.

Spots are limited and filled on the strength of the work. We read the applications. This is a selective, no-cost research program, and we treat it as one.

The Commitment

A partner is counting on the team. Every Resident owns a real part of the work that the others depend on. That shared accountability is part of what makes the experience worth it.

• Show up to every session, prepared.
• Do the work between sessions.
• Stay accountable to your team and the cohort each week.
• Follow through on what you take on, across both semesters.

What Partners Get

As a partner, you come with something you want to bring AI into: a research problem, a workflow, or a project. We put a team on it for two semesters, in a consulting-like arrangement. We treat you as a client: the team works independently, under a SoTS Residency Mentor, toward a result you can use. The students stay ours, working your problem, not joining your group.

The need can be specific, such as a model for a particular task or analysis code made reproducible, or open-ended: "I want to use AI somewhere in my research workflow." The team takes it from there. They scope the problem, test what AI can do for it, and build toward a result you can use.

The team covers the full stack, from your science to the implementation. They work out which modern methods fit the problem: data science and visualization, classical machine learning, deep learning, and today's generative and agentic AI. They engineer the result into reproducible, usable software. You need no technical background of your own.

Your time commitment is light: at most an hour every two weeks, often less. The team does the work and brings findings back to you.

The team is a small group of serious students, selected on demonstrated research and first-principles reasoning, and placed by role to fit the problem. They bring real effort, a careful perspective on what AI can do, and the persistence to follow a question past its first obstacle.

Every team is supervised by Dr. Sergey Samsonau, who works on both sides of this intersection: he guides research teams, and he applies AI to science himself.

He has spent a decade helping scientists bring modern tools into their work: teaching Essentials of Data Science to PhD students and postdocs at Columbia in 2016, serving as AI Technical Lead at NYU for five years, and leading research-enabling AI and data science initiatives across disciplines. He developed the methodology this program runs on.

Dr. Samsonau organized the AI Meets Science conference, the first of its kind for the NYC metro area, with participants and top voices in the field from NASA, NIST, the Simons Foundation, SandboxAQ, Columbia, Cornell, Princeton, NYU, Rutgers, Fordham, Arizona State, and Stevens Institute of Technology. We invite and work with partners from across that community: universities, national labs, government science agencies, and companies.

The Residency is a nonprofit initiative, offered through the Authentic Research Access Program at NOPI, a registered 501(c)(3). You contribute an open question to a public-interest program.

Because the program is educational and public-interest, the engagement can count toward the broader-impacts and outreach components that grant proposals ask for, such as NSF Broader Impacts. Ask us for a short paragraph you can include in a proposal.

Want to bring AI into your research? Write to us at team@teenscientists.org.

For Donors

The Residency is free for every admitted student. It is offered through the Authentic Research Access Program, a nonprofit initiative fiscally sponsored by NOPI, a registered 501(c)(3). The program expands access to authentic research education and builds public-interest research infrastructure.

Donations keep seats open for capable students regardless of ability to pay and support open research records anyone can inspect. To give, visit the Authentic Research Access Program at NOPI.