Daily small groups

Your home group for the week

The 40 participants are split into 8 fixed home groups of five. You stay in the same group all week, so you come to know your groupmates' projects as well as your own — and each member facilitates exactly one day.

Why fixed groups

What the small groups are for

Every content day (Days 1–5) has two small-group blocks: one mid-morning to apply the day's topic to your own project, and one at the end of the day to step back and talk through how your thinking moved. You spend them with the same home group all week.

The groups are fixed on purpose. The aim is not to match people by subfield — quite the opposite. Each group is deliberately mixed across research areas, methods, institutions, and countries, so that over the week you give and get feedback from people who see your problem from angles you wouldn't have considered. By the end of the week you should understand each of your groupmates' projects nearly as well as your own.

Each day's discussion has a job: refine your project against the day's topic — problem framing, then data, evaluation, methods, and a Day 5 stress-test. The group is where a lecture stops being abstract and starts changing your project.

Groups are mixed by design, not by topic. Yours spans different research areas, methods, institutions, and countries on purpose — so don't expect your group to share your subfield. Work on whatever you came to work on; the value is the outside perspective.

How the sessions run

What happens in a small group

These are working discussions, not quiet work time — the goal is to think out loud about your project and engage with everyone else's. There's just enough shared structure to get going; the rest is open, and the day's facilitator steers it.

  • Day 1, first block — introductions. Introduce yourselves, then give a short pitch of the project you came with. Keep it brief on purpose: if it doesn't land in a quick description, that's useful signal about how the problem is being communicated — better to surface it now than to over-explain.
  • Days 1–4, morning block — work the day's prompt. Start by walking through that day's Project Workbook prompt together. You won't have prepared it in advance — that's the point; use the time to work out how the day's topic applies to each project.
  • Days 1–4, closing block — how your thinking moved. Save the last ~30 minutes for a round where each person takes about five minutes on how their thinking about their project evolved over the day, anchored on the Workbook.
  • Day 5 — present, then red-team. Each person presents their updated project in brief, constrained form (around 12 minutes) — problem, data, evaluation, and method — and the group then red-teams it as a whole: assume it's built and deployed, and find where and why it would fail in the real world. Use what surfaces to scope risks or revise your framing.

Outside those anchors the discussion is open. The Project Workbook is the shared reference, but it's a private thinking aid, and you're encouraged to keep refining it after each day.

The rotating facilitator

Facilitating a small group

There is no instructor in the room — the groups are peer-led. Because each group has five members and the week has five content days, you each facilitate exactly one day, leading both of that day's small-group blocks (the mid-morning and the closing one). The facilitator is not the expert; your job is to keep the discussion useful and make sure every project — and every person — gets time. Beyond that, run the session however you think works best. In particular:

  • Keep time. Divide each block so every project gets a roughly equal turn — about 15 minutes each in a group of five. Don't let the first project eat the hour.
  • Get everyone talking. Make sure each member speaks — about their own project and about others'. The goal is discussion, not silent work; draw in anyone who's gone quiet, and make it clear everyone is both invited and expected to contribute.
  • Hold the day's focus. Keep each turn anchored on the day's topic (framing, data, evaluation, methods, or — on Day 5 — red-teaming), and gently pull the conversation back when it drifts.
  • Push for the concrete. Steer toward specific, usable feedback and a clear next thought for each project — not vague praise.

On the other four days you're a participant: come ready to talk through your project and to give your groupmates sharp, generous feedback. If you can't make the day you're assigned, swap with another member so every day still has a facilitator.

The 8 groups

The groups

Find your group below. Each has an identity color — it matches your name-tag sticker — and a constellation. Members are listed in the order they facilitate: the first leads Day 1's discussions, the next Day 2, and so on through Day 5. Days 6 (job talks) and 7 (posters) have no home-group session.

Group 1 · Orion

In facilitation order — each member leads the day shown.

  • Jiho Kim Day 1 LLM agents for error-free electronic health records
  • Tae Jones Day 2 Human-centered AI for health equity
  • Divyam Madaan Day 3 Multimodal learning for smarter medical data collection
  • Reyhaneh Hosseinpour Day 4 LLM virtual patients for clinical training
  • Amy Tai Day 5 Privacy-preserving AI for cancer diagnostics

Group 2 · Lyra

In facilitation order — each member leads the day shown.

Group 3 · Cygnus

In facilitation order — each member leads the day shown.

  • Guilherme Imai Aldeia Day 1 Interpretable AI for brain-disorder diagnosis
  • Daeun Kyung Day 2 Reinforcement-trained doctor agents
  • Ferdaous Idlahcen Day 3 Computational pathology for gynecologic oncology
  • Ben Fox Day 4 Foundation models for sleep and physiological signals
  • Ha Le Day 5 Adaptive wearable AI for human activity recognition

Group 4 · Aquila

In facilitation order — each member leads the day shown.

Group 5 · Perseus

In facilitation order — each member leads the day shown.

  • Yvonne Wu Day 1 Generative AI for student mental well-being
  • Haoran Zhang Day 2 Fairness and robustness in clinical machine learning
  • Dominik Becker Day 3 Machine learning for 3D medical imaging
  • Simon A. Lee Day 4 World models for health AI
  • Jiyoun Kim Day 5 Multimodal EHR modeling and clinical AI evaluation

Group 6 · Andromeda

In facilitation order — each member leads the day shown.

Group 7 · Cassiopeia

In facilitation order — each member leads the day shown.

Group 8 · Draco

In facilitation order — each member leads the day shown.