REAL-WORLD SUCCESS
The client, a government-operated service branch, was tasked with optimizing the maintenance schedules of its large fleet of vehicles. To solve this issue, the client’s transportation contractor sought assistance to enable accurate predictive maintenance.
Faced with the challenge of integrating two key datasets—parts eligible for maintenance and their maintenance time estimates—the absence of a direct link between the datasets meant the client could not automatically project time estimates for maintenance identified at the part level, posing a significant obstacle in forming schedule optimizations. While this task could have been done manually, it would have taken a significant number of staff hours per vehicle.
StandardData was tasked with unifying these datasets to enhance operational efficiency and accuracy from one data source.
StandardData’s innovative approach to data integration challenges and a desire for a scalable solution resulted in leveraging open-source technology.
StandardData employed a HuggingFace sentence transformer model to link part names with maintenance task titles semantically. The joins now take two computer hours to complete.