Overview
In the ever-evolving maritime industry, optimizing voyages to improve operational efficiency, reduce fuel consumption and carbon emissions are strategic priorities. Syroco, a deeptech leader in maritime decarbonization, partnered with Automat-it to strengthen its machine learning operations (MLOps) capabilities using AWS technologies.
This collaboration enabled the deployment of a robust, scalable, and automated pipeline to support the continuous training, evaluation, and deployment of digital twins powering route optimization and vessel performance analysis.
The Challenge
As Syroco expanded its fleet optimization services, the company needed an MLOps framework capable of supporting high-frequency model updates, rigorous evaluation, and robust version control. These capabilities were essential to maintain the performance of the models used in digital twins, in dynamic maritime environments and to scale deployment across vessels and customer contexts.
The Solution
AWS-based MLOps Platform
Automat-it supported Syroco in designing and implementing a production-grade MLOps infrastructure, built entirely on AWS. The solution focused on industrializing the model lifecycle with an emphasis on automation, reproducibility, and scalability.
1. Model Training and Experimentation
- Amazon SageMaker Pipelines: Automates the end-to-end ML lifecycle, from preprocessing to training and evaluation
- SageMaker HPO (Hyperparameter Optimization): Boosts model performance through automated tuning
- Deep Neural Networks (DNN): Used to capture complex relationships in maritime operational data
- MLflow: Ensures full traceability of experiments, datasets, and model versions
2. Model Deployment and Monitoring
- SageMaker Model Registry: Manages multiple model versions, streamlining deployment workflows
- Amazon CloudWatch & SageMaker Model Monitor: Enable proactive monitoring of model behavior, including performance and data drift detection
- GitHub Actions: Integrates CI/CD processes to secure and automate model deployment and updates
3. Observability, Governance, and Cost Efficiency
- MLflow & TensorBoard: Track training performance metrics across experiments.
- AWS Auto-Scaling & Spot Instances: Optimise resource utilization and reduce infrastructure costs
- Role-based access control (RBAC): Enforces security and governance across development and production environments
The Results
This MLOps pipeline has become a cornerstone of Syroco’s AI infrastructure, enabling:
Reliable and Repeatable ML Workflows: Ensuring consistent deployment of high-performing models
Faster Model Iteration: Supporting continuous improvement through streamlined retraining and evaluation cycles
Robust Monitoring: Providing transparency and control over model behavior in production
Scalable Infrastructure: Facilitating the deployment of tailored models across vessels and customer-specific contexts
The collaboration between Syroco and Automat-it led to the successful implementation of a production-grade, AWS-based MLOps platform. This foundation empowers Syroco to accelerate innovation in maritime optimization while ensuring reliability, efficiency, and scalability — all key to supporting its mission to decarbonize maritime transport through advanced technology.
