SNU Connectome Lab · PI Jiook Cha
2026 Winter Undergraduate Research Internship
Help build the NeuroX autonomous research agent, an AI researcher meant to form and test its own hypotheses with brain foundation models. Three undergraduates (any major) will spend eight weeks, 4 January – 26 February 2027, building one module that goes into a lab paper, from implementation to the results table. The lab plans to extend the agent from human to mouse and monkey brain data, and students with a deep interest in animal neural data are especially welcome, as are students planning graduate study in 2027 or 2028. We look for students who push through hard problems until they find the best solution, and for students who reframe a given problem into a more important question.
The full announcement, including the three research tracks and the learning task, is in Korean. 공고 전문 (한국어)
Research project: the NeuroX autonomous research agent
The lab is building the NeuroX autonomous research agent. The goal is to build an AI researcher that uses brain foundation models as tools to turn a research question into a hypothesis, design the experiment (analysis) that tests it and record it in advance (pre-registration), analyse the data, check for itself whether the result is statistically trustworthy, and learn and reason from the result to revise its next hypothesis. Brain data are growing fast, yet rigorously testing one hypothesis still takes months. The aim is not just a fast AI but one that guards against its own false discoveries, and that handles human, mouse and monkey brain data in the same loop.
First versions of the research-loop skeleton, the statistical self-check, the pre-registration record and a connection to one brain model exist, and this winter the interns help extend them. The three tracks are core modules of one agent: T1 is its eyes for reading brain data, T2 widens its view from humans to mice and monkeys, and T3 is its judgement, doubting and checking its own results. Each intern builds and tests one core module in one track; assignment is decided after selection, based on interest and readiness.
- T1. The agent's eyes: human brain models as tools (fMRI, EEG). Connect the lab's fMRI and EEG foundation models as tools the agent can call, and make them refuse mismatched input. The deliverable is one brain model connected as an agent tool and a results table verifying it on valid and invalid inputs.
- T2. Brains across species: mouse and monkey data. Connect public mouse (calcium imaging, Neuropixels) and monkey (spike recording) data to the loop and test open questions such as whether representations learned in one species transfer to another, or whether anatomical knowledge actually helps the model. The deliverable is a results table from a first cross-species comparison or knowledge-injection experiment. Students choose the species, and work starts from public datasets. Students with a deep interest in animal neuroscience or neurophysiology data are especially welcome, even with little PyTorch experience.
- T3. An AI scientist that checks itself. Let a language model run the research loop (hypothesis, experiment design and pre-registration, analysis, check) and measure how much the statistical self-check prevents false discoveries. Applicants with little brain-imaging or deep-learning implementation experience can apply.
PyTorch experience is recommended for T1.
What selected interns receive
- A personal NVIDIA DGX Spark account from week 2
- Research AI API usage (OpenRouter and similar) covered by the lab without a usage cap, for research purposes
- A stipend of KRW 600,000 per month before tax
- Weekly mentor meetings, lab meetings and PR-based code review
Selection timeline (KST)
- 20 Nov (Fri), 18:00 KST
- Application deadline (documents, learning task, online survey)
- 25 Nov (Wed)
- Screening results; research challenge and personal AI API key sent
- 25 Nov – 7 Dec (Mon), 23:59
- Research challenge (pre-registered prediction due 1 Dec)
- 11 Dec (Fri)
- Interview invitations (about 2–3× the 3 positions)
- 14–18 Dec
- 30-minute in-person interview at SNU
- 22 Dec (Tue)
- Final offers; programme starts 4 Jan 2027
How to apply
Email a CV, an application statement and the learning task (two A4 pages) to connectome@snu.ac.kr, and complete the online survey, by 20 November 2026, 18:00 KST.
Subject line: [2026동계인턴지원] 이름_학교_진학희망연도
Details of each document are in the Korean announcement.