Build a real AI agent — in one day.
A hands-on, in-person hackathon at the DevRev office in Palo Alto. High school and undergraduate students team up to build a working AI agent for computational oncology or quantum-AI systems biology — with mentors, prizes and a stage.
Welcome to the QxBio 2026 Hackathon
A hands-on, in-person event at the DevRev office in Palo Alto where you will build a working AI agent on Computer by DevRev to tackle a real challenge in computational oncology or quantum-AI systems biology. Below is what to expect, what to bring, and what is expected of you.
1. Who can take part
High school students
The Explorer tier is for high school students. Framing and communication weight more in scoring.
Undergraduate students
The Builder tier is for undergrads. Design and technical depth weight more in scoring.
Age & parental consent
18+ complete their own consent. Under-18 must submit parental consent, a liability waiver and a separate photo/video consent before attending.
First-come, first-served
Registration is capped for the venue. Apply early; your application stays under review until all required documents are submitted.
2. Teams
2 – 4 members
The event runs about 12 to 15 teams. Register as a pre-formed team or as an individual.
One team per person
Individuals are grouped into teams by tier, theme preference and complementary skills on the day.
Mixed-level teams
A mixed-level team competes in the tier of its most senior member.
Team captain
Each team names a captain who owns the roster, the submission and the presentation order.
Lock at 9:40 AM
Team and challenge choices are locked at 9:40 AM on event day. Changes after that need Event Director approval.
3. What you will build
Each team picks one theme and one problem statement.
AI Agents for Computational Oncology
- A research assistant
- Patient-journey or triage helper
- Lab-operations agent
- An explainability companion
AI Agents for Quantum-AI Systems Biology
- A quantum-concept tutor
- Research-coordination agent
- Simulated hardware operations
- Literature-to-hypothesis agent
Your primary agent must be built on Computer by DevRev, in your assigned workspace. Any other tool, external model, dataset or pre-existing code you use must be disclosed. You will work only with the synthetic or approved public datasets provided. Real patient data, protected health information, confidential records and personal credentials are strictly prohibited.
Scope test: a good one-day project completes one narrow workflow reliably. A broad idea without a testable result scores poorly even if the topic is impressive.
4. What to bring
Charged laptop
Bring a compatible charger. Install or update a current browser if instructed.
Approved account email
And your school / student ID if requested.
Accessibility / dietary notes
Anything not already submitted through your registration form.
Login test before the day
Complete the Computer by DevRev login test with the exact email you registered with. Don't bring real patient data or a pre-built submission.
5. Your submission
Every team submits the following. Submit a title / team / problem stub by noon; the final form closes at 1:15 PM.
A working demonstrator
A working demonstrator in your assigned workspace.
Problem, user, approach
Problem, user, approach, data, test result, limitations and next step.
Workflow image + notes
An architecture or workflow image plus a short README / reproducibility note.
External-tool disclosure
Any external models, tools, pre-existing material or assistance used.
3-minute live demo
Show the workflow, not slides-only. Questions follow.
Show your thinking
- A user story
- A visual explanation
- At least one test case
Show your engineering
- A functional workflow
- An original method
- Evidence from at least two cases where feasible
File-naming: use QXB26_[TIER]_TEAM##_ProjectName — letters, numbers and underscores only. Late acceptance requires a logged organizer-controlled technology incident. After the freeze, only demo-stability fixes are allowed — no new features or content.
6. How your day runs
7. Working with mentors
Guidance, not building
Mentors guide with questions and critique — they won't build, debug, operate or rewrite your agent. The work must be yours.
Three checkpoints
Confirm scope and success test → diagnose a reasoning, data or workflow block → pressure-test evidence, limitations and the demo.
Open, observable spaces
All mentoring happens in open areas. A minor is never alone with one adult in a closed room.
Use the Tech Desk
Send account, platform and dataset problems to the Tech Desk rather than using mentor time.
8. Where to get help
Account, login, platform or data
Challenge interpretation or scope
Team conflict or inactive member
Consent, accessibility, welfare or conduct
Submission, room order or scores
Emergency
9. Venue and access rules
Designated hackathon zone
Stay within the assigned area of the DevRev office — no other floors or restricted areas without an escort.
Visitor badge at all times
Follow the parent drop-off / pickup and sign-out arrangements.
Minors — approved sign-out only
Minors are released only through the approved sign-out process.
10. Conduct and integrity
Respect
Treat people, property, venue rules and staff directions with respect. Harassment, discrimination, intimidation, retaliation and deliberate disruption are prohibited.
Authorship & honesty
Do not misrepresent authorship, results, data sources, limitations or outside help.
Approved accounts & data
Use only your approved accounts and datasets; never share passwords or upload confidential material to any AI tool.
Photography
Photography and publication respect your stated media choice and the separate photo / video consent for minors.
Enforcement
Organizers may warn, remove or disqualify for unsafe conduct or integrity violations.
11. Recognition
Every team is reviewed
A panel of expert judges reviews every team.
Prizes for top submissions
Top submissions in each tier receive prizes.
Everyone gets recognised
Every participant receives a certificate or letter of participation.
Guidelines for Judges
Thank you for judging QxBio 2026. Your role is to give every team a fair, evidence-based and consistent assessment across both tiers and both themes.
Scoring criteria & tier weights
| Criterion | Explorer | Builder |
|---|---|---|
| Problem Framing and Relevance | 25% | 20% |
| Agent Design and Reasoning Quality | 20% | 25% |
| Technical Execution on Computer | 15% | 20% |
| Usefulness / Real-World Impact | 15% | 20% |
| Presentation and Communication | 25% | 15% |
Rating anchors (1 to 5)
Missing, materially incorrect or unsupported; no credible evidence.
Partial attempt with major gaps, weak evidence or unclear logic.
Clear and functional at the expected tier; adequate evidence and limitations.
Well reasoned, reliable and useful; good evidence; minor gaps only.
Distinctive, rigorous and compelling for the tier; excellent evidence, judgment and communication.
Additional guidance
Your role and the panels
Calibration and timing
What to look for
- Problem Framing: a precise problem and intended user; scope fits the day; data and constraints match; importance is justified.
- Agent Design: a coherent workflow; tool and agent choices are justified; reasoning handles uncertainty and failure.
- Technical Execution: the agent functions in the sandbox; data is integrated and disclosed safely; testing and stability are evident.
- Usefulness: output is actionable for the stated user; benefits and risks are realistic; limitations and failure modes acknowledged.
- Presentation: a clear, on-time demonstration with visible evidence; the pitch is not used for background only; questions answered directly.