Know exactly when to clean your hull.
We turn your vessel's existing operational data into precise cleaning recommendations — using physics-informed AI, not calendar guesswork.
The cost of biofouling is real.
These numbers are from your vessels, not hypothetical scenarios.
Source: IMO 2023 Biofouling Guidelines · Schultz et al., Biofouling, 2011 · Ocean AI internal validation
Biofouling is costing you money every day.
Marine organisms attach to your hull within weeks of launch. By the time a cleaning is scheduled, you may have lost months of fuel efficiency. Calendar-based maintenance is not evidence-based — it cleans too late, too early, or both.
Fouling severity vs impact
| Level | Fuel penalty | Speed loss |
|---|---|---|
| Light | 5% | −0.5 kn |
| Moderate | ~10% | −1.0 to −1.5 kn |
| Heavy | 20%+ | −2.0+ kn |
IMO 2023 Biofouling Guidelines; Schultz et al., Biofouling, 2011
Calendar-based maintenance
Fixed intervals ignore actual vessel condition. You clean when the schedule says — not when the data says. Some cleanings happen too early (wasted cost), others too late (wasted fuel).
Data-driven maintenance
Ocean AI monitors your hull condition continuously. Clean only when the net present value of a cleaning becomes positive. Every intervention is justified by the numbers.
From sensor data to a cleaning decision.
Continuous data collection
We connect to your vessel's existing sensor streams — shaft power, GPS speed, draft, wind. No new hardware required.
Physics + AI performance analysis
We compare your actual speed against the speed a clean hull should make in the same conditions — the speed gap. This separates genuine fouling from weather and loading, and converts it directly into extra fuel burn.
Optimal cleaning window
We calculate precisely when the net present value of a cleaning becomes positive — when savings justify the intervention.
Actionable recommendation
You receive a clear recommendation with expected savings, payback period, and confidence level. Data-driven, not calendar-driven.
Physics + AI. Not a black box.
Pure data-driven models cannot reliably distinguish a fouling signal from a prediction error caused by unusual weather or loading. Our architecture encodes the physics of ships as hard constraints — so the AI only corrects what physics leaves unexplained, reading lost speed and extra fuel burn together and telling weather from fouling, hull from propeller.
Technical overviewModel structure
Speed prediction
We predict the speed a clean hull should make from the physics baseline — hull resistance, propeller–hull balance, and wind & wave loads — then add a small, physics-constrained correction for the structured effects physics alone cannot explain, like loading changes or fuel-quality variation. The gap that remains is the fouling signal.
Condition estimate
The gap between expected and actual speed feeds a condition tracker that reports hull state with a calibrated confidence range — fused into a single maintenance alert score, so you act on evidence rather than a single noisy reading.
Physics Baseline
The physics of the ship, built in
The speed a clean hull should make is derived from first principles — hull resistance, propeller–hull balance, and wind & wave loads — not a single simplified rule. This keeps predictions physically consistent even in conditions never seen during training.
Neural Correction
Physics-constrained learning
A compact learning layer accounts only for the structured effects physics cannot explain — loading changes, fuel-quality variation, operational transients. It reads zero on a clean hull, so any sustained drift in the speed gap is real degradation.
Condition Monitor
Calibrated confidence
Every estimate comes with a confidence range, not just a number. Several independent checks watch the speed gap — and the system escalates only when multiple lines of evidence agree.
Shaft power, GPS speed, draft, wind, IMU — existing data streams
Physics-based speed model — the speed a clean hull should make in these conditions
How far actual speed sits below a clean hull — zero when the hull is clean
Hull condition with a calibrated confidence range — not just a single number
Multiple checks must agree — over 99% of single-sensor spikes are ignored
Clean now / Watch / Alert / Schedule — with expected savings and confidence
Research-grade science. Production-grade engineering.
TU Delft Spin-off
Research founded at Delft University of Technology
YES!Delft
Member of the YES!Delft innovation network
AINed Maritime AI InnovationLab
Track 2 (Emission Reduction) consortium member
2 Vessels Validated
Results validated on real operational data from two vessels
The right package for your fleet.
Every plan pays for itself. No new sensors. No hardware. Start from one vessel.
Assess
Know where you stand.
per vessel · one-time
- Fuel & emissions cost of past fouling
- Cleaning effectiveness analysis
- Degradation rate benchmark
- Full report + data export
Monitor
Never miss the optimal window.
per vessel · per month
- Continuous real-time condition tracking
- Watch → Alert → Recommend signals
- Optimal cleaning window + payback period
- Uncertainty-quantified recommendations
Scale
Operate at fleet intelligence.
5+ vessels · tailored contract
- Everything in Monitor
- Trim optimisation module
- Fleet-wide benchmarking
- Custom API integration + SLA
Built by maritime AI researchers.
Andrea Coraddu
Co-founder
TU Delft
Luca Oneto
Co-founder
Simone Minisi
Co-founder
Olena Karpenko
Co-founder
Ready to stop guessing?
Clean at the right time. Save fuel. Reduce emissions. Improve profitability.