At a glance
A major utility client wanted to proactively maintain power line components to minimize wildfire risk and improve service.
Our client was facing high maintenance costs due to manually inspecting millions of images to identify defective components in their power lines. They wanted to automate this process and enable preventative maintenance, all without putting undue stress or changes on their internal teams.
Approach and solution
We knew that the use of AI was a cost-effective strategy to automate the identification of defects using computer vision models and a highly scalable machine learning (ML) platform. The client had a data science team in place, but needed help streamlining productivity and building the ML platform. This required analysis of existing systems and the introduction of a new plan to improve them.
We began our approach by conducting assessments to understand the client’s existing team capabilities and data architecture. The results of these assessments allowed us to create recommendations for team training and development of the platform. We built two teams: a data science team and a machine learning engineering team.
With the data science team, we established a center of excellence for their data science practice by:
Meanwhile, with the ML engineering team, we started to develop the ML platform by:
Value and benefits - “the wins”
The client’s data science teams now use established best practices to build highly accurate computer vision models and to efficiently deploy them in production. This enables the utility to eliminate manual inspections and preventatively maintain components of their power lines.