Sign up to know more
Operationalizing Machine Learning Workflow with Microsoft Azure
The client deals with planning, design, and construction management of water and wastewater-related projects – from clean water treatment, storage, and distribution to wastewater and stormwater collection, treatment, and reuse.
The client had developed an algorithm to predict the speed of water streams on an hourly basis to plan resources in their water treatment plan. They faced issues with the algorithm that was sitting in a silo, preventing multiple data scientists/developers from working on it together and provided a challenge in creating multiple versions of their algorithm. Additionally, they were looking to capture model performance in development environment and to automate their machine learning workflow using Azure for orchestration of production data.
The client required a data model which would take care of model development, model tuning, model versioning, and model deployment. In addition to the above data platform the client wanted a scheduled job to run on an hourly basis to fetch data from Azure SQL database, pass the data as input to the deployed model, retrieve the predictions and load them in Azure SQL database.
To cater to client needs, SNP leveraged Azure Machine Learning Service for deploying client algorithm on to the cloud and Azure components to deliver a cost-effective and scalable data model platform and hourly job.
The client has started using the Data Factory pipeline (scheduled as an hourly job) to generate Power BI reports to cater to their BI needs and to plan resources in their water treatment plant by looking at the predictions coming from the model sitting on Azure.
