Prepare a reproducible project.

Register and qualify first

Every participant must individually register, fork, clone/run the A/B/C baseline, document architecture and sanitized qualification evidence, and be selected before building. Model D knowledge is not required.

Choose your challenge

• Smart Login Guardian
• Find the Anomaly
• API Traffic Watch
• Explain My Security Alert
• Predict the Outage
• AI for Everyday Life

Submission fields

Review stages: Received → Pending review → Verified, Rejected or Withdrawn. Verification does not grant publication permission.

Prepare public-safe repository, exact commit, offline setup/run/test commands, synthetic inputs and actual results. The original five challenge briefs were inspected locally; AI for Everyday Life is a Portal extension pending research COM-003A alignment. Official Google Forms have not been changed. View the research repository ↗. Never create a public issue with private contact information.

Individual entry and teams

• Individuals may compete; teams may contain two to four participants, maximum four.
• Every member, including the leader, must individually complete qualification and be selected.
• Designate one team leader and obtain a unique team ID from the organizer.
• No duplicate active team membership; membership changes require organizer approval.
• Organizers verify rosters in private Google Sheets; roster mappings are never public.

Mandatory qualification and selection →

Final project submission

Every project must be a working MVP, executable prototype or functional proof of concept. A slide deck or documented walkthrough alone is insufficient. Any programming language or framework; standalone implementations are acceptable after mandatory AeglysAI qualification.

• Team ID
• Selected challenge
• Public GitHub source URL and frozen review commit/tag
• Project README
• Reproduction/setup instructions
• Working demonstration video URL — maximum two minutes
• Explanation of AI/ML or intelligent technique
• Test or validation evidence
• Data/model attribution and licensing
• Known limitations

Narrated screen recordings or silent captioned demonstrations are accepted. Webcam footage is not mandatory. A deployed demo URL is optional. Submit only sanitized evidence; no private participant, credential or patent-related data.

Two-round judging

Round 1 — technical screening

Review source code, working prototype, required video and reproducibility. Round 1 determines finalist selection.

• Functionality — 30 points
• AI/ML or intelligent feature — 25 points
• Relevance and creativity — 20 points
• Reproducibility — 15 points
• Video clarity — 10 points

Round 2 — finalist review

Shortlisted finalists give a five-minute live demonstration over Google Meet followed by five minutes of technical Q&A. Round 2 determines final ranking; do not combine round scores.

• Live demonstration — 35 points
• Technical understanding — 25 points
• Solution quality and safety — 20 points
• Technical Q&A — 15 points
• Presentation — 5 points

Proposed capacity: five to ten finalist teams, subject to submission quality and organizer capacity. This is not a guaranteed finalist count. Each rubric totals 100 points; equally weighted, non-conflicted judges score independently and their scores are averaged within each round.

Round 1 ties: functionality, intelligent feature, then reproducibility. Round 2 ties: live demonstration, technical understanding, then solution quality and safety. Remaining ties receive equal additional demonstration/Q&A and independent panel review; unresolved ties receive a shared rank with any capacity decision explained under published rules.

Reviewers disclose conflicts privately and recuse; organizers assign unconflicted replacements. Technical clarification uses equivalent time and scope, is logged privately and cannot permit undisclosed post-deadline feature development. Scores, roster mappings and private comments remain in private Google Sheets/Drive. Judges, dates and results are not confirmed.