What is a Hackathon?
1-Minute Summary for Beginners
“A Research Time Machine: Complete 3 months of research in 3 nights and 4 days”
A Hackathon is a combination of “Hacking” and “Marathon.” Here, hacking doesn’t mean cybercrime - it means “the joy of efficiently and technically diving into difficult problems to solve them.”
Think of it like an Iron Chef competition:
- Regular Research: You go shopping for ingredients, wash them, prepare them, and spend months pondering recipes.
- Hackathon: You bring perfectly prepared premium ingredients (data) and recipes (hypotheses), and the best chefs (researchers) focus insanely on cooking (coding/experiments) for 3 nights and 4 days in an isolated space to create unprecedented results.
We block out daily distractions (emails, administrative work) and focus 100% on research and development. By compressing processes that would normally drag on, we aim to produce tangible outcomes (models, code, paper drafts) that would be impossible alone in a short time.
Why is Preparation Critical?
For a 40-person, 5-lab Hackathon Success
This hackathon isn’t a small study group - it’s a large-scale project with 40 participants from 5 labs. Coming empty-handed with a “let’s figure it out there” attitude will result in 100% failure. Especially with Neuro-AI data that is massive and requires long training times - starting preparation on-site means you’ll spend all 3 nights and 4 days just transferring data.
For a successful hackathon, we strongly request completion of 3 essential preparations before departure.
✅ Essential Preparation 1: Research Topic and Hypothesis Definition
The first day of the hackathon should be when you run your first code, not when you decide on a topic. Prepare a “testable hypothesis,” not just a vague idea.
- Bad: “I want to analyze EEG data with deep learning.” (too abstract)
- Good: “To verify Emotion Contextualized Perception, develop a new Transformer model on the ABCD dataset and produce benchmarking scores within 3 days.” (specific)
- Action Item: Each team must complete a 1-page proposal with [research goal, expected outcomes, model structure] before the kickoff meeting.
✅ Essential Preparation 2: Data Curation and Preprocessing [Most Important]
Neuro-AI data (fMRI, EEG, ECoG, ABCD Dataset, etc.) is massive and complex. You cannot afford to waste two days downloading 26TB of data or removing noise in the on-site internet environment.
- Data Acquisition: Finalize the dataset for analysis and obtain access permissions for secure data (ABCD, etc.) in advance.
- Preprocessing Complete (Model-Ready): Do NOT process Raw Data on-site. Complete noise removal, format conversion (BIDS, etc.), and dimensionality reduction beforehand, and prepare data in a form ready for immediate model training (Train_X.npy, etc.) on external hard drives or cloud storage.
- Environment Testing: Test in advance that required libraries, GPU server access, Docker environment, etc., are working properly.
✅ Essential Preparation 3: Sufficient Brainstorming and Role Assignment
The on-site is for execution, not discussion. To prevent 40 people from wandering aimlessly, each person’s position must be clear.
- Role Definition: Decide who will do core coding (Core Modeler), who will handle data pipelines (Data Engineer), and who will handle result interpretation and visualization (Analyst).
- Scenario Check: Discuss even “If this model doesn’t work, what is Plan B?”
Sample Daily Schedule
| Time | Activity | Expert Tip |
|---|---|---|
| 09:00 | Breakfast & Daily Stand-up | “How did last night’s training go?” Share briefly and get straight to work. |
| 10:00 | Sprint [Morning Focused Coding] | Your brain is clearest now. Write the hardest code during this time. |
| 13:00 | Lunch | |
| 14:00 | Co-Work [Afternoon Research & Mentoring] | Don’t struggle alone - ask the PhD student in the next room. Hackathons are about collective intelligence. |
| 16:00 | Tea Time, Walk, Break [Mandatory] | Take your eyes off the monitor - new solutions emerge. |
| 18:00 | Dinner | |
| 20:00 | Sprint [Late Night Development (Optional)] | Manage your energy wisely. Set long-running model training during this time. |
Expert Recommendations for Successful Hackathons
1. Tangible Outcomes
“Working hard” doesn’t matter. Aim for visible outputs like “trained model weights,” “validated benchmarking design,” “working prototype.”
2. Collaboration over Competition
The reason 5 labs gathered is for synergy, not competition. Actively absorb other labs’ know-how (GPU optimization, data interpretation perspectives) and break down barriers between labs through Hybrid Parallelism.
3. Documentation (Power of Records)
Brilliant ideas and code from the hackathon will be forgotten when you go home. Record all progress and troubleshooting processes in real-time to preserve them as research assets.
Only those who are prepared can experience the miracle of 3 nights and 4 days.
Come fully armed with data and ideas, and we’ll see you at the hackathon site.
— 2026 Neuro-AI Hackathon Organizing Committee