Canvass recently announced its latest update to the Canvass AI platform, which dramatically improves industrial engineers’ ability to extract value from their data and advance operations. We sat down with Jeff Wood, Canvass AI Director of Product Management, to learn more about what is included in this new release and what is coming up.
Jeff Wood: We are very excited about this new release as it builds on our vision to empower industrial engineers, through AI, to extract value from their data and create real impact across their organization. This release, Canvass AI version 4.1, focuses on enabling the core user journey, including data ingestion, data exploration, data preparation, model training, and model evaluation, so that the user can complete all of these steps in a single platform.
Canvass AI v4.1 includes capabilities that focus on solving three key challenges:
· Knowing when a model is ready to deploy
· Cutting lengthy data preparation cycles
· Scaling AI
We focused this release on the core user journey to enable users to quickly become proficient with AI techniques and capabilities to augment their process and operational knowledge. The Canvass AI platform makes it easy for users to harness their data, accelerate their digital transformation journey, and build AI capability across their company.
Deciding when a model is ready for deployment can be difficult. We’ve made it easier by adding features that help users understand how an AI model is working and why it generates the predictions it is generating. Explaining AI is important because it ensures a user is confident that their AI model will generate accurate predictions for their operations or helps them identify areas in which the AI model needs further iteration.
The newly added Compliance Score, for example, allows the user to determine when a model is ready to be put into production by showing the percentage of predicted values that fall within a user-defined threshold.
The Error Distribution Chart provides a view into the range and distribution of errors between predicted and real values, allowing users to compare models and determine which is ready to deploy.
Steve Lohr of The New York Times once said that 80% of a data scientist’s time is spent on data wrangling and data prep. We’re working to flip this statistic on its head with three key features. The first allows users to convert the data type right within the platform, such as numerical versus categorical entry. It’s an automated feature that eliminates a labor-intensive and mundane task.
For one happy customer, this has saved a week in data prep time.
The second feature is the Time Interval Removal. This feature allows the user to exclude certain time periods that they don’t want their model to be trained on, such as a planned maintenance period or when the plant was closed. This allows the user to train the model with data more representative of normal operating conditions, which will produce a higher quality model.
The Data Curation Visualization feature saves users’ time in the model training process by providing a preview of how data preparation selections have impacted the dataset before training the model. This will significantly cut the model training cycles and time by enabling users to make sure the data selections are right before they get into the training process.
AI pilot purgatory has become common across the industrial sector. Canvass AI is built to overcome this challenge by embedding a guided AI journey across the platform to make it easy and fast for industrial companies to go beyond one use case and transform into an AI organization.
As an example of how we are helping our customers have greater success on their AI journey, we have added capabilities that create greater efficiency and remove the time-consuming obstacles. This includes the ability to clone a project, which is useful if they have data in one project that they would like to use in another project - saving the user the time and hassle of re-uploading data.
This new release is currently available to all existing customers. The Canvass AI platform is a cloud-based software-as-a-service (SaaS) solution, so for existing customers, there is nothing that they need to do. In addition, to make learning new features easier, we have included an on-screen guide that helps users better understand the new capabilities.
We are very excited about what the future holds for Canvass. We will continue investing in the platform for industrial engineers to continue to gain value every time they bring in data and then scale those capabilities across different boilers, pumps, users, teams, plants, processes, and outsets. We have already started working on our next release, Canvass v4.2, targeted for release in October. With this release, we will add even more capabilities to empower the user on their AI journey, extract value from their data, and expand the application to new use cases.
