Enabling an Entertainment & Leisure Company to Operationalise Machine Learning at Scale

Services provided

Artificial Intelligence
Machine Learning
MLOps
Cloud Engineering

Platforms used

AWS (Amazon SageMaker Pipelines,
SageMaker Model Registry,
SageMaker Endpoints, SageMaker Batch
Transform, AWS Service Catalog, Amazon S3)

Engagement length

8 weeks

Other stats

2 machine learning models productionised
Model deployment automated across development, test, and production environments
Daily model performance monitoring implemented
MLOps accelerator deployed, providing the customer with a reusable foundation for
future projects
Model onboarding time reduced from approximately 1 week to 1 day
Manual retraining effort reduced by approximately 2 hours per retraining cycle

Background on the customer

Our customer operates a portfolio of entertainment and leisure experiences that rely on forecasting and predictive analytics to support operational planning and business decision-making.

The organisation had invested in machine learning experimentation and developed predictive models to support business operations. However, the models remained isolated within notebook-based environments, limiting their ability to generate real-time insights and deliver sustained business value.

Challenge

The organisation had successfully developed machine learning models for occupancy forecasting and demand prediction but faced challenges moving these models from experimentation into production.

The existing approach created several limitations:

Machine learning models were maintained within Jupyter notebooks
Predictions relied on static datasets and manual execution
Forecasting workflows required several hours of manual effort
No standardised model deployment process existed
Limited visibility into model performance and drift over time
Lack of governance, lineage, and approval controls
Difficulty scaling machine learning across additional use cases

Without a production-ready MLOps capability, the organisation risked ongoing operational inefficiencies, inconsistent prediction outputs, and reduced return on its machine learning investments.

Stack Highlights

Amazon SageMaker Pipelines Amazon SageMaker Model Registry Amazon SageMaker Endpoints Amazon SageMaker Batch Transform AWS Service Catalog Amazon S3 AWS CloudFormation AWS IAM

The Approach

Codex designed and delivered an AWS-native MLOps platform that automated the end-to-end machine learning lifecycle from model training through to deployment and monitoring.

Our approach included:

1. Assessing existing machine learning workflows and deployment challenges

2. Productionising occupancy and demand-forecasting models

3. Setting up MLOps accelerator in the customer’s AWS account

4. Implementing SageMaker Pipelines to automate model training, evaluation, and packaging and establishing a governed model approval process using SageMaker Model Registry

5. Creating automated deployment pipelines across development, test, and production environments

6. Setting up both real-time and batch inference patterns based on business requirements

7. Introducing automated performance monitoring and model-drift detection

8. Packaging the solution as a reusable Service Catalog offering to accelerate future machine learning initiatives

This approach created a repeatable and governed framework for operationalising machine learning across the organisation.

Technical Outputs

Codex delivered a production-ready MLOps capability that transformed machine learning from isolated experimentation into a scalable operational platform.

Key outputs:

Productionised two machine learning forecasting models

Automated model training, evaluation, approval, and deployment workflows

Established a governed model lifecycle using SageMaker Model Registry

Implemented real-time inference through SageMaker Endpoints

Enabled batch forecasting using SageMaker Batch Transform

Established environment promotion processes across development, test, and production

Introduced automated model performance monitoring using business-relevant accuracy metrics

Implemented drift detection and threshold-based retraining processes

Created reusable infrastructure and deployment patterns through AWS Service Catalog

The resulting platform provides a scalable foundation for future machine learning initiatives while reducing operational overhead and increasing model reliability.

QA Platform Highlights

Production-Ready MLOps Framework
Transforms machine learning from isolated notebook-based experimentation into governed, repeatable production workflows.

Automated Model Lifecycle Management
Automates training, evaluation, approval, deployment, and monitoring across multiple environments.

Continuous Performance Monitoring
Provides ongoing visibility into model accuracy, drift, and degradation, enabling proactive model management.

Reusable Machine Learning Foundation
A Service Catalog-based deployment model enables new machine learning projects to be onboarded quickly using standardised governance and controls.

Scalable AWS-Native Architecture
Built on Amazon SageMaker services to support future machine learning use cases without additional platform redesign.

Business and Commercial Outcomes

The MLOps platform transformed the organisation’s ability to operationalise and scale machine learning initiatives.

Reduced manual retraining effort through automated model lifecycle management
Improved reliability and consistency of machine learning predictions
Enabled continuous monitoring of model performance and drift
Reduced new machine learning project setup time from approximately one week to one day
Improved governance through model versioning, approval workflows, and lineage tracking
Increased confidence in production machine learning outputs
Established a repeatable framework for future AI and machine learning initiatives
Accelerated the transition from experimentation to operational business value

The platform now enables the organisation to deploy, monitor, and scale machine learning solutions more effectively while providing a governed foundation for future AI innovation.