Thomson Reuters Accelerates Use of Machine Learning to Drive Deeper Business Insights
Enabling machine learning for all
Thomson Reuters (TR) engaged an AWS Resident Architect (RA) to help design an Enterprise Artificial Intelligence (AI) Platform. TR had been working for several years to develop AI capabilities to manage a wide variety of complexities, such as customer churn, financial forecasting, and various complex Natural Language Processing use cases. The AI Platform was started by TR Labs, who provided secure access to the cloud for data scientists. TR wanted to enhance their AI Platform and provide self-service capabilities for business and engineering teams. Specifically, TR wanted their business teams to apply machine learning (ML) frameworks and address business challenges at scale and speed. TR also wanted to standardize their multi-account, multi-persona ML environment for all of their business units and empower data scientists to deliver ML models faster, while still adhering to evolving governance standards.
Enterprise AI Platform extends ML capabilities through a “single pane of glass”
The RA helped TR validate their AI Platform strategy and architectural design. The RA worked with multiple teams, including enterprise architecture, data engineering, data science, operations, and business. The RA helped TR design a loosely-coupled micro-service architecture to build an ML lifecycle starting with experimentation, model registry, inference deployment, and custom metric monitoring components. All these components are now coupled with automation and governance. Upon completing the design for the AI Platform, TR was able to initiate a build process to accelerate “minimum viable products” to production.
I love working with our Resident Architect. She's a trusted partner building the AI Platform. She excels at breaking problems down into their individual components and finding solutions quickly to then move to implementation. She helped bridge a gap in our team that we would not have been able to do on our own.”
Vice President of AI/ML & BI Platforms, Thomson Reuters
Accelerating innovation through AI
In the span of the RA engagement, TR built and tested their first ever Enterprise AI Platform. Teams across TR can now develop and build ML models faster and also migrate legacy ML models. At the end of the RA engagement, TR identified multiple ML models for migration to the AI Platform, with a goal to have most models managed by the following year. Using the ML Platform capabilities in Amazon SageMaker, TR business and engineering teams can securely process gigabytes of data. Using advanced insights from their ML models, TR is able to address wide-ranging business challenges such as customer churn, financial forecasting, and speech analytics. At the same time, TR Labs continues to push Amazon SageMaker boundaries for product-enabled AI solutions.
“TR teams can now use AWS capabilities to train and release an ML model within weeks, compared to months," says Apazoglou.
AWS Services Used
- Amazon SageMaker for experimentation, training and inference
- MLOps strategy using Amazon SageMaker model registry, Amazon SageMaker Pipelines, AWS CodePipeline and AWS CloudFormation
- Monitoring capabilities using Amazon SageMaker Model Monitor, extended with customization
- Extended model governance metadata with Amazon SageMaker model registry
About Thomson Reuters
Thomson Reuters is a leading provider of business information services. Our products include highly specialized information-enabled software and tools for legal, tax, accounting and compliance professionals combined with the world’s most global news service – Reuters. For more information on Thomson Reuters, visit tr.com and for the latest world news, reuters.com.
About AWS Resident Architect
The AWS Data Lab Resident Architect program provides AWS customers with guidance in refining and executing their data strategy and solutions roadmap. Resident Architects are dedicated to customers for 6 months, with opportunities for extension, and help customers (Chief Data Officers, VPs of Data Architecture, and Builders) make informed choices and tradeoffs about accelerating their data and analytics workloads and implementation.