Vessel-specific voyage optimization with machine learning: Manta on AWS

This article describes how Manta Marine Technologies built a cloud-native data infrastructure on AWS to deliver vessel-specific voyage optimization using machine learning models trained on each ship's operational history.

• Vessel-specific ML models trained on individual ship data outperform generic vessel-type averages
• Apache Airflow on EC2 orchestrates data pipelines and ML workloads across ECS and Lambda
• Data stored in RDS and DynamoDB with security enforced through least-privilege IAM and KMS encryption
• Graviton-based instances provide 25-30% better price performance for ETL pipelines
• Continuous retraining loop improves recommendations as new voyage data accumulates
• Expected fuel savings of 2-4% per vessel through optimized voyage guidance

Manta's AWS-based solution enables ship operators to receive personalized voyage guidance that continuously learns from each vessel's performance, improving operational efficiency and reducing costs.