SeaGuard HeatWatch · 3-minute judge guide
Heat-stress early warning for port and offshore workers in Abu Dhabi. Machines detect, fixed rules recommend, AI only explains: a person always decides.
Click a step: each opens the right screen, already set up. No sign-in needed. Use the top bar to come back here.
📂 Source code and README on GitHub · Backup demo links: Google Cloud Run · Microsoft Azure (Azure may take about 30 s to wake up)
- Open ▸
A worker on a warm morning
One level, one action, a water reminder and the Abu Dhabi ADOSH zone. Big buttons: "I feel unwell", "I need a break".
≈ 30 s · Worker - Open ▸
Afternoon heat: Abu Dhabi Withdrawal zone
45 °C, calm air: Stop work. ADOSH rule shown: emergency work only, under permit, 20 min work / 40 min rest, 1.2 L water per hour.
≈ 30 s · Worker - Open ▸
Man down, in Arabic
The simulated watch plays a fall with no movement: the face turns red. Only "level 4, possible man down" leaves the watch; no heart rate or age.
≈ 35 s · Worker watch - Open ▸
Supervisor: no plain "OK"
Crews at High or Stop need a real action. Press "+5 min" twice: an unanswered alert goes to the site manager. Adherence is measured.
≈ 40 s · Supervisor - Open ▸
Control centre: human in command
Site map of the port, live watches, man-down to answer within 3 minutes, and a log of every human decision.
≈ 45 s · Control centre - Open ▸
Guided demo and benchmark (extra)
Seven simulated workers replayed minute by minute: supervisor board, worker band, anonymous health office, private wellbeing check-in. Ends with the 52-case benchmark: 15/15 RED caught, 0/16 false alarms vs 8/16 for WBGT alone.
≈ 60 s · Guided demo
Try more
Where to see each assessment criterion
| Criterion | Where to look |
|---|---|
| Health impact and clinical relevance | Hydration and kidney protection on every level; ACGIH strain limits on the watch (steps 1, 3) |
| Technical depth and architecture | WBGT + Abu Dhabi TWL, the stricter wins; body heat estimated from heart rate on the watch (steps 2, 3) |
| Working solution and demo quality | Everything on this page runs live; 73 automated tests |
| Responsible AI, safety and data governance | AI never sets a level; no plain "OK"; body data stays on the watch; every decision logged (steps 3, 4, 5) |
| Scalability and deployment readiness | Any watch brand, any cloud, 12 languages, works offline with a printed fallback |
| Commercial viability and path to pilot | Pilot targets and pay-on-results in the deck (slide 11) |
- Workers, crews, watches and positions are synthetic demo data. The back-test uses real Abu Dhabi 2025 weather (Open-Meteo, CC BY 4.0).
- All non-English texts are AI-drafted translations, to be checked by native speakers.
- TWL is our own estimate, to be validated against the official ADOSH calculator.
- Not a medical device. Thresholds will be reviewed with clinical mentors.