TABLE OF CONTENT

About the Client

ETP is a leading retail software provider headquartered in Singapore, serving market leaders across 24 countries in the Asia Pacific, India, and the Middle East. With over 500 enterprise software projects, ETP powers 300+ brands and 35,000+ stores with its robust omnichannel retail solutions. The company specializes in providing AI-driven insights, real-time analytics, and seamless customer experiences, making it a key technology partner for retail businesses worldwide.

Business Situation

ETP faced significant inefficiencies in managing AI/ML model deployments across multiple tenants:

Manual deployment processes – AI/ML models had to be configured manually for each tenant.
High error rates – Manual setups led to inconsistencies and operational inefficiencies.
Scalability bottlenecks – Deployment processes were unable to support rapid business expansion.
Delayed onboarding – New tenant integration was slow due to labor-intensive configurations.
Security concerns – Lack of automation increased risks in deployment security and compliance.
Limited real-time insights – Prediction models required faster processing and retrieval mechanisms.

The Solution

Unthinkable automated ETP’s AI/ML model training and deployment, ensuring a seamless, error-free, and scalable process:

To overcome the challenges of manual deployment, Unthinkable implemented an automated AI/ML pipeline using Apache Airflow and MLflow. This system scheduled weekly model training and deployment while dynamically onboarding new tenants with minimal human intervention. The anomaly detection model was optimized with an extended isolation forest algorithm, while the recommendation engine was designed to provide personalized recommendations in real time. The infrastructure was hosted on Google Cloud Platform (GCP) to ensure high performance and scalability.

Automated AI/ML model deployment to eliminate manual setup inefficiencies.
Dynamic tenant onboarding with auto-configuration and real-time activation.
Scalable infrastructure built using Airflow, MLflow, and GCP for seamless expansion.
Enhanced prediction accuracy with real-time anomaly detection every 20 minutes.
API-driven recommendation system for user-specific content personalization.
Security-first architecture with isolated backend environments for each tenant.

The Impact

By automating AI/ML model training and deployment, Unthinkable significantly improved ETP’s operational efficiency. The onboarding time for new tenants was reduced dramatically, while predictive insights became more accurate and timely. The automation also ensured higher security, minimized deployment errors, and provided a scalable framework for future AI enhancements. This success led ETP to entrust Unthinkable with another critical project—automating their forecast model.

Conclusion

Unthinkable’s AI/ML automation solution transformed ETP’s deployment process, enabling faster onboarding, improved accuracy, and scalable operations. By eliminating manual interventions and ensuring real-time AI-driven insights, ETP strengthened its position as a leading retail technology provider. The success of this collaboration paves the way for further innovations in ETP’s AI ecosystem, ensuring continuous improvements in efficiency and customer experience.