Productionalize Machine Learning Models for HVAC Industry using MLOps

Challenge

A company in the global HVAC (heating, ventilation, and air conditioning) industry, serving tens of thousands of customers worldwide, faced a significant challenge in utilizing advanced data from their appliances.

These appliances continuously generate sensor data and metadata every second. So the company had developed two patented machine learning models, one focused on energy savings calculations and another on detecting leaks and monitoring the overall health of specific appliances.

However, their main challenge was deploying these models into production to fully leverage the data and improve operational efficiency.

Solution

Algorhythm Group set up the Azure Machine Learning platform (AML) in combination with a third party feature store (Feast) and other Azure services to meet the monitoring capabilities.

So, all the required environments, compute resources, pipelines, model registries and endpoints were prepared and deployed to AML.

At the same time we followed the necessary guidelines to ensure reliable, efficient and good model building practices.

Impact

Thanks to our Azure Machine Learning platform, the two patented machine learning models are now available in production. This achievement allows for time reduction in future model development.

At this point our client now has all the tools and processes in place to perform MLOps. Finally they can efficiently serve/track models to serve business and consumer needs in an agile way.