01
Business needs
One of the major American airlines sought a cutting-edge AI/ML solution to streamline the document verification process for customer travel readiness. They wanted to:
- Automate the approval process through the Travel Readiness Center (TRC)
- Enhance customer travel experience and reduce the workload on gate and contact center agents
Automated validation of 4.5M+ documents, saving 191K agent hours
02
Solution
The Impetus team migrated the application from on-premises .NET to AWS cloud. We developed a generic framework to support multiple models with a unified endpoint and standard input to address the airlines’ imperative for an advanced AI/ML model.
The team crafted a versatile OCR Model Inference Process, leveraging a generic framework to support multiple models through a unified endpoint. Hosted on Amazon EKS Fargate containers, the custom-developed Rest APIs in Go overcame API-Gateway limitations, ensuring scalability through Horizontal Pod Autoscalers (HPAs).
The solution incorporated extensive reporting and monitoring capabilities, including ad-hoc and detailed reporting options. Data was ingested into Amazon Athena and Elasticsearch through an AWS Glue pipeline, supporting visualization and dashboards for funnel analysis. Additionally, an OCR Audit Tool, a microservice application deployed on Amazon EKS/AWS Fargate, facilitated manual auditing with visual UI for document and model output comparison.
Highlights
- CI/CD automation: Enabled one-click deployment with TeamCity, AWS CodePipeline, and GitHub integration for streamlined iterative development.
- Security measures: Implemented robust KMS encryption, IAM Roles, and code scanning for compliance with Veracode, Qualys, and Wiz tools.
- Model execution reports: Integrated into a UI for document/transaction searches using AWS Lambda functions and AWS Glue jobs.
- Monitoring and alerting: Utilized Datadog integration for comprehensive monitoring and alerting capabilities.
- Scalability: By leveraging AWS EKS, we achieved horizontal scalability and manage the workload
30% increase in operational efficiency with automated model deployment
03
Impact
The AI/ML solution revolutionized the airlines’ document validation process. More than 4.5 million documents were automatically validated, resulting in:
- Efficiency gains: Reduced manual effort for gate and contact center agents, enhancing passenger experience
- 30% increase in operational efficiency with automated model deployment
- Saved 191K agent hours
- Automated ML model deployment: Implemented with the latest technology stack, enabling seamless iterative development
- Model monitoring: Established automated model monitoring and drift calculation using Elasticsearch/Kibana dashboards.
- OCR audit tool: Facilitated faster manual auditing of documents with a visual UI, promoting swift decision-making
- SSO integration: Enabled Single Sign-On leveraging OAM and ALB/Cognito authentication for seamless onboarding of new agents in the OCR Audit Tool application
- 30% increase in operational efficiency with automated model deployment
