Play hard. Play fair.

A short list, intentionally. Anything not on it, you can probably do — but ask a mentor if you're not sure.

01 · Teams. Up to four members. All members must be currently students (any major). Teams may form up through the opening ceremony — team-formation help is available at check-in.

02 · Pillars. Pick one. Stay there. Each team picks exactly one pillar and works that pillar's assigned dataset. You cannot switch pillars after the build window opens.

03 · In-person. At least two team members must be physically present at Events on Pine during the 13-hour Friday build window. Remote work is permitted after 10:30 PM Friday.

04 · Tools. Pretrained models: yes. Closed-API LLMs (Claude, GPT, Gemini): allowed in your application layer only — they do not count toward your Model score. You must train or fine-tune at least one model yourself. Open-source fine-tuning (HuggingFace, torchvision, PyTorch, etc.) is fair game.

05 · Data. External data is a supplement, not a replacement. You may bring in external datasets to augment the assigned one. Disclose any external sources in your Model Report.

06 · Attribution. Stand on giants. Cite them. Open-source code and tutorials are fair use with attribution. Do not copy another team's repo. When in doubt, cite it in the README.

07 · Feature choices. You decide what to keep. Document the choices in the Feature Choices and Trade-offs section of your Model Report. "We didn't think about it" is not an answer. Judges will probe this in Q&A.

08 · App behavior (SML/USML). Exploration, not approval. The Bank Marketing and Credit Card Applications apps must be exploration tools — for the agent, for the applicant, for the user. No team's app may render an auto-approve, auto-deny, auto-call, or auto-skip verdict on a person.

09 · Code of conduct. Treat fellow participants, organizers, judges, and volunteers with respect. No harassment, discrimination, or disrespectful behavior of any kind. Mentor support is available throughout the build window — ask for help early, not in the last hour.

Disqualification. Plagiarism, submitting another team's work, using pre-built project code from before the event, misrepresenting team member contributions, violating dataset terms, or violating the code of conduct will result in immediate disqualification.