Predikto: Max
Background
Predikto Max is a predictive analytics tool used by data scientists at Raytheon Technologies to configure and run machine learning to accurately predict failures for large equipment (such as jet engines) before they occur. Since Raytheon Technologies' acquisition of Predikto in 2018, Max had been rebuilt from the ground up to ensure it could meet the needs of a large-scale, defense-focused enterprise such as Raytheon. While significant improvements had been made to the back-end architecture of the application, gaps in functionality and inefficiencies had been introduced for users by a rush to quickly rebuild the UI and adoption of third-party tools (such as Airflow and Kibana) to augment functionality that was previously native to the product.
Approach
Over roughly 12 months, this project involved leading a small team of designers and researchers to conduct ongoing research to identify significant gaps in functionality and pain points for data scientists working between Predikto Max, Airflow and Kibana. Work was focused on prioritizing and addressing these areas for improvement and benefited from close collaboration with our Product and Project Management, and Software Engineering counterparts on our team.
Outcomes
- UI rebranding and adoption of Material-UI React framework to align with Predikto Product Suite and Raytheon Technologies guidelines and improve overall product look and feel.
- Successful implementation of a Predikto Max component library, standardizing common UI components, reducing front-end dev rework, and improving front-end serviceability and UX fit and finish overall.
- Successful delivery of features improving efficiency for configuration retrieval and management in Predikto Max (eg searching, tagging and sorting).
- Successful implementation of linking strategy to improve efficiency when navigating between related objects in Predikto Max, Airflow and Kibana and reduce functional overlaps and potential risks associated with integration dependencies.
- Successful implementation of versioning and naming conventions that allow users to identify the parent of a child object in Airflow and Kibana - the reverse of the linking strategy.
- Development of a notification/alert interface for end users to receive predictions (in progress).
