The Setq Challenge
3 tracks. One mission: build AI that works
in regulated procurement — and prove it.
What is it
AI built for the real world.
Not the demo stage.
The Setq Challenge is setq.ai's open competition for engineers, researchers and founders who believe AI should solve problems that matter — regulated procurement, sovereign infrastructure and industrial supply chains.
We focus on measurable, verifiable performance against real benchmarks — not investor decks or polished prototypes.
All winning solutions must be open sourced. We believe progress toward applied AI in defence and industry should be shared with the community that needs it most.
- ✓ All authored code must be open sourced (MIT or CC0)
- ✓ No proprietary LLM APIs during final evaluation
- ✓ Solutions must run on EU infrastructure only
- ✓ Results must be reproducible from your repository
- ✓ Teams of 1–5 people, any nationality
- ✓ Submissions via GitHub + benchmark dataset
Competition tracks
3 tracks. No shortcuts.
Choose one track. Build something real. Open source it.
Procurement Intelligence
Build a system that reads complex RFQs, identifies part numbers and NSNs, sources compliant suppliers and produces a fully auditable proposal — faster and more accurately than a human team on the same case.
Sovereign Data Pipeline
Design and demonstrate an AI pipeline that processes sensitive procurement data entirely within EU infrastructure — no US cloud, no third-party APIs. Every action must be logged, traceable and governance-compliant.
Open Track
Any AI application to regulated industrial, maritime or defence operations that demonstrates a measurable improvement over the current human baseline — with real data, reproducible benchmarks and a fully open-sourced solution.
Timeline
Key dates for 2026
Competition opens
Benchmark datasets released. Applications open.
Milestone #1
Interim leaderboard published. Top 10 teams notified.
Milestone #2
Final shortlist. Finalists invited to Demo Day.
Submissions close
Final submissions + open source repos due.
Winners announced
Winners announced. Partnership opportunities revealed in Lisbon.
Evaluation
Judged on what actually works.
No points for impressive decks. All tracks are evaluated on verifiable, reproducible performance against the benchmark dataset. Judges cannot override the numbers.
Speed
End-to-end processing time vs. human baseline on the same case set.
Accuracy
Correctness of outputs — part numbers, supplier selection, compliance mapping — against verified ground truth.
Compliance
Auditability of the pipeline. Every action must be logged and traceable to a regulation or certification requirement.
Sovereignty
Data residency, isolation and governance by design. Evaluated by independent infrastructure audit.