Waylay ML Ops enables customers to effectively operationalize any machine learning algorithm with the Waylay rules engine.

Waylay ML Ops is for data scientists, analysts and engineers that are working on data solutions involving ML model training. Once an ML algorithm is trained and ready to produce output predictions, it can be used with Waylay ML Ops as a simple node in business logic workflows, unlocking multiple automation and integration applications.

When is machine learning used in the IoT domain?

Machine learning can be used whenever large amounts of data are available and is especially useful in industrial IoT for processing image data, in visual object recognition. Many enterprises are realising that placing modern cameras in legacy environments can be a minimally intrusive and efficient way of kick starting digital transformation.

A rich pool of raw visual information is collected from cameras placed on machines on production lines, cameras placed on worker helmets or drone cameras, to name just a few. New value can be captured by training ML models on these camera feeds to ultimately achieve improved production quality, worker safety or maintenance procedures, among others.

Execute any machine learning model with Waylay

The Waylay BYOML (Bring Your Own Machine Learning) module enables customers to effectively operationalise any machine learning algorithm with the Waylay rules engine

After training, the first level of data processing is done by running the ML model on the raw image data, to obtain the predicted answers to the preset question(s) that the model was trained to answer. Questions like:

  • is the product as per defined parameters (colour/size/layout etc.)?
  • is the product missing any parts?
  • is the installation configuration correct (e.g. is the valve in the correct position?)
  • is the installation safe? (e.g. is there corrosion detected?)

Once in possession of these answers, companies can use them to trigger new data workflows that follow specific business logic. This is where Waylay BYOML comes in.

Why would you execute your ML model with Waylay?

The reason why you would want to use the ML model inside an automation and integration platform such as Waylay is because of the extra value that the model can bring to the business by making it operational.

Being able to detect whether the goods on your production line are up to standards through visual inspection is very important. Being able to automatically take customized actions depending on what the visual inspection model finds, is an even more exciting achievement.

These actions can be:

  • alerting machine operators
  • sending notifications to maintenance teams
  • stopping / starting machines
  • creating / updating entries in ERP systems
  • initiating any business processes in any back-end IT system

Why bring your own ML?

Data science teams have different preferred frameworks that they work with. Waylay ML OPs supports popular ML frameworks such as TensorFlow, Scikit-learn, Keras, PyTorch and allows you to call the model either:

  • from your preferred external service, like AWS Sagemaker or Azure ML or
  • from Waylay, after uploading it from a Jupyter notebook for example.

Once your model is hosted, you will be able to use it as a sensor node inside the business logic, such as in the example below, where you can see a rule in the Waylay drag-&-drop rule designer, that includes an Azure ML-hosted model. We have presented this demo-use case in this blog article, where you can also see a recorded video demo of how we set-it up and how it runs live.

The model’s output is interpreted according to the business logic of this demo assembly line, and two different alarms are triggered depending on whether a fault in the process is detected or whether a specific machine is underperforming.

Learn more

With Waylay ML Ops companies can orchestrate their machine vision models by using the output of ML models to build automation workflows.

To learn more about using machine learning in Waylay, visit the technical documentation page and watch our on-demand webinar on how to operationalize machine learning.