Assessment Services Ideas Portal

Recent experiments using the EM AI software have shown that the current step of primary analysis can be replaced by an AI marking of the data with no reduction in final report accuracy, and an estimated 10% decrease in labour required. The experiments are being performed with no software integration, involving a human to markup the model results behind the scenes.

To put the model into production, we need to either integrate it into the existing Pipe Vision software. Note that this would apply to all PCCP analysis, not just that from Pipe Diver. Ideals cannot be attached to the general domain of PCCP analysis.

Components include:

• Create a generic way to for the data scientists to publish AI models with self-describing metadata

• Publish the models using the current newly created framework.

• Update the analysis software to be able to call into the generic AI model

Development costs estimated at 1 month of data science to get the model architecture ready and 3.5 months of software to integrate it into the application $56K CAD

Michelle estimates efficiency of the new workflow being 1-2 days per project.