MantisGrid AI Hackathon 2026

The 2026 MantisGrid AI Hackathon ("the Hackathon") is organized by MantisGrid AI Inc ("the Organizer") and runs from September 17, 2026 from 9:30am to 3:30pm (PDT), in person in Palo Alto, California. By registering, you agree to these Rules and the Code of Conduct (Section 10). If anything here is unclear, or you're unsure whether something is allowed, ask the organizers by emailing hankunli@mantisgrid.ai — asking first is never penalized.

Participants must be 18 or older to be eligible for prizes. Organizers, judges, mentors, employees of MantisGrid and event sponsors, and their immediate families may not compete, but are welcome as mentors, volunteers, and judges. 

• Teams of up to 5 people. Each person may join only one team.

• Team rosters may be changed (add/remove/swap members) until 24 hours before the event. After that, rosters are locked.

• Every listed team member must have materially contributed to the project.

• All judged work must be created during the Hackathon. You may arrive with an idea, a plan, or even a project you've worked on before — but you may not reuse pre-existing code you wrote, and if you build on an existing project, judges will only consider functionality added during the event.

• Open-source libraries, frameworks, starter kits, and the provided starter notebooks and templates are all fair game, provided you comply with their licenses and meaningfully build on them. A thin wrapper or reskin of an existing tool (including vllm-semantic-router or the OTel demo used as-is) will score poorly on Innovation and may be disqualified from awards.

• Submit one project to one track. You may combine ideas across tracks; if you do, tell us which track's judging focus you want applied.

• When time is up, stop building. Bug fixes and deployment repairs after the deadline are fine; new features are not.

Participants may use any AI model, large language model (LLM), AI coding assistant, agent framework, or open-source AI tool during the hackathon. This is an AI infrastructure event — using AI heavily is expected, not penalized.

In exchange, we require transparency:

• Your repository README must list the AI models, coding assistants, and agent frameworks you used, and briefly state what was AI-generated versus built by the team.

• Judges may ask any team member to explain any part of the submission. Inability to explain your own architecture and code will hurt your Technical Execution score and may trigger review.

• Misrepresenting AI-generated work as hand-built, or another team's/person's work as your own, is grounds for disqualification.

• Track resources (public infrastructure datasets including the starter notebooks, routing templates, sample workloads, OTel sample data) are provided under their respective licenses and terms; you are responsible for complying with them.

• Any API keys issued for the event are per-team, must not be shared or committed to your repository, and will be revoked after the event. Usage may be rate-limited; abusive usage (e.g., using event keys for non-hackathon purposes) is grounds for disqualification.

• Do not use datasets containing personal data you don't have rights to, proprietary data from an employer, or data obtained by scraping in violation of a site's terms.

• OpenTelemetry track note: keep AI agents in a human-assist role — explaining, exploring, and suggesting — rather than claiming autonomous failure prediction. Projects making unsupportable reliability claims will be marked down on credibility under Technical Execution.

Submit via Devpost before September 17, 2026 3:30pm PDT. Late submissions are not judged. A complete submission includes:

1. Team information (all members, with student/career status).

2. Project title and description, including your chosen track.

3. A public source code repository — with API keys and secrets removed — including the AI-disclosure README from Section 5 and instructions to run the project.

4. A demo video of 2 minutes or less showing the project actually working. Slides-only videos (without live demo) don't qualify.

Submissions must be in English and must not contain content that violates the Code of Conduct.

Projects are scored by a judging panel against the published criteria: Technical Execution (40%), Innovation / Wow Factor (30%), Potential Impact (20%), Presentation / Demo (10%). Track-specific judging focus (data storytelling; routing logic and explainability; usefulness on a real standard) is applied within these criteria.

Awards: Best Overall Project, Most Innovative, Best Business Impact, Best User Experience. Each team is eligible to win at most one award. 

Using MantisGrid AI is not required to compete or win the four general awards — the 10% relevance criterion can be satisfied through infrastructure concepts alone.

Ties are broken by the Technical Execution score, then by judges' vote. Judges' decisions are final and binding. Organizers may verify any submission (including running the code and reviewing commit history) before confirming winners.

• Your project remains yours. All submissions remain the intellectual property of the participants who created them.

• By submitting, you grant the Organizer a non-exclusive, worldwide, royalty-free license to use, display, and reference your submission (including the demo video and screenshots) for judging and for promoting this and future Hackathons, for 3 years from submission.

• You warrant your submission is your original work (per Section 4), that you have rights to everything in it, and that it does not infringe any third party's copyright, trademark, patent, privacy, or publicity rights.

All attendees — hackers, mentors, judges, sponsors, volunteers, and organizers — must follow the MLH Code of Conduct, at the venue and in all event channels. It also applies to project content. Report incidents to Hankun Li hankunli@mantisgrid.ai; reports may be made anonymously. Have FUN — that's a rule too, but the Code of Conduct is how we make sure everyone gets to.

Organizers may, at their discretion, disqualify a team for violations including but not limited to: breaking these Rules, breaking the Code of Conduct, plagiarism or misrepresentation of work, misuse of provided API keys or data, or other unsporting behavior. Where practical, teams will be given a chance to respond before disqualification is final.

• Prizes are non-transferable and may not be substituted except by the Organizer. Winners are solely responsible for any applicable taxes. Team prizes are split equally unless the team tells us otherwise.

• The Organizer and sponsors are not liable for damages, losses, or injuries arising from participation, to the maximum extent permitted by law. 

• The Organizer may amend these Rules before the event begins with notice to registered participants. After hacking begins, no material changes to eligibility, judging criteria, or the award structure will be made except to correct errors, and any change will be announced to all participants.