AI-Powered Maritime Intelligence

How Atracai Works

We analyze publicly available AIS data with advanced machine learning to predict port congestion and berth availability. No complex integrations. No port cooperation required.

The Core Insight

Port congestion and berth availability follow predictable patterns. By analyzing millions of historical vessel movements, we can forecast when berths will become available - without needing access to port systems.

Queue lengths follow patterns
Berth turnaround times are predictable
Vessel size affects wait times
Seasonal trends are learnable
7.2M+
Vessel snapshots analyzed
Training our prediction models
11M+
AIS positions
55+
Berths mapped

The Prediction Process

From AIS signal to actionable prediction in four steps

1

Data Ingestion

We continuously ingest AIS data for all vessels within 350 nautical miles of target ports.

Real-time AIS feeds
Updates every 15 minutes
350nm coverage radius
2

Feature Extraction

Raw AIS data is processed to extract meaningful signals about port state and vessel behavior.

Vessel classification
Zone detection
Queue analysis
3

ML Prediction

Our models, trained on millions of historical data points, generate predictions.

Time-to-berth prediction
Congestion forecast
Confidence intervals
4

Delivery

Predictions are delivered through our dashboard and API for your operational use.

Live dashboard
REST API
Alerts (coming soon)

Our Data Sources

All publicly available - no special access required

AIS Data

Automatic Identification System - mandatory broadcast from all commercial vessels over 300GT.

  • Position, speed, course
  • Vessel identity (MMSI, IMO)
  • Destination (when set)

Vessel Registries

Public databases with vessel characteristics that inform our predictions.

  • Vessel type and size (DWT)
  • Length, beam, draught
  • Flag and classification

Historical Patterns

Years of historical data reveal patterns that inform our predictions.

  • Past port call durations
  • Seasonal congestion trends
  • Berth turnaround times

Continuous Validation

We don't just make predictions - we validate them against real outcomes

How We Measure Accuracy

Holdout Testing
Models trained on historical data, tested on unseen data
Live Tracking
Every prediction is tracked against actual berth arrival
False Alarm Detection
Vessels heading elsewhere are identified and filtered

Current Accuracy Metrics

0-8 hour predictions 1.7h MAE
8-24 hour predictions 3.8h MAE
24-48 hour predictions 5.8h MAE
Congestion forecast (24h) 90% AUC
See live validation dashboard

Why No Port Integration?

Traditional JIT solutions require complex integrations. We don't.

Traditional JIT Approach

  • Port system integration

    Requires access to port management systems

  • Terminal cooperation

    Every terminal must agree to share data

  • VTS data feeds

    Vessel Traffic Service agreements

  • Months/years to deploy

    Complex stakeholder alignment

Atracai Approach

  • Public AIS data only

    Globally available, no permissions needed

  • No port cooperation

    Works independently of port systems

  • ML learns from patterns

    History reveals future behavior

  • Works today

    Start receiving predictions in days

See Our Methodology in Action

Explore our live dashboard to see real predictions being made and validated.