Setting Up MLOps Platform to Provide Insights on Emissions

The Challenge

In response to new emission policies for maritime trades, the project operationalized ML models using MLOps methodologies and tools to provide emission insights and predictions, allowing for better advising of ship owners, charters, and operators on major emission regulations.

The Approach

In this project, the cloud architecture was designed according to the specific requirements.

Terraform scripts were written to establish the cloud environment, while a CI/CD pipeline was set up and pre-processing code was containerized. Batch prediction pipelines were created with Vertex AI, alongside retraining pipelines.

Finally, an online prediction infrastructure was established with authentication, and TabPy was utilized within Google App Engine.

The Result

Eraneos was able to sucessfully simplify the setup and maintenance of cloud infrastructure by utilizing Terraform scripts.

They also implemented an automatic CI/CD pipeline triggered by code changes to update software artifacts.

Additionally, the team established automatic batch prediction based on a customer-defined schedule, as well as automatic predictions based on user input.

Furthermore, they were able to automate the retraining and monitoring of ML models, providing improved efficiency and accuracy.