One platform for all your users

We have built Criterion AI from the ground up to make it easy for all of your users to leverage the power AI—whether they are specialists, operators, QA managers, project leaders or directors.

At its core, Criterion AI handles three key workloads required for achieving success with AI: managing data, training, and validating models and deploying them to production.


Data management

We understand that you may have a lot of data stored across a wide variety of IT systems in your organization. And we respect that. Therefore, we have made it incredibly easy to access your data no matter where they may be stored. Either use our web interface to easily upload your data or put our FTP(S) interface to work.

In addition, we have built a set of incredibly easy-to-use tools to enrich and annotate your data, which may be required for some models.

Learn about how to manage data in Criterion AI →

Training models

Using the data you uploaded to Criterion AI, you can train state-of-the-art deep learning models for tasks such as image classification, image segmentation, anomaly detection, predictive maintenance and a whole lot more.

Our model zoo hosts a bunch of preimplemented models that you can start training with your own data right away. In addition to that, we let you build your own custom models using Python and a deep learning framework like Keras or TensorFlow. Check out our docs on custom models.

Learn about how to train models in Criterion AI →

Validating models

We take validation very seriously—both in terms of the general regulatory requirements for IT systems (as covered by 21 CFR Part 11 and ISO/IEC 27001) and in relation to ensuring that the performance of the models we train can be relied upon. While the use of artificial neural networks in pharmaceutical manufacturing is still a new topic, some regulation does exist on the use of multivariate data analysis (e.g., Ph.Eur. 5.21 Chemometric Methods Applied to Analytical Data, USP General Chapter <1039> Chemometrics as well as ICH Q8–12).

Criterion AI ensures that model validation takes place in compliance with all of the relevant regulatory requirements, which means that you can focus more on building great models and less on all of the boring paperwork.

Learn about how we validate models in Criterion AI →

Deploying to production

Once you have trained and validated your model, you can deploy it into production with just a few simple clicks. Choose whether you want to deploy your model in the cloud or in an on-premise environment. Online deployments get a RESTful interface, which makes it super easy to interact with them from other applications. Online deployments can also be used directly from Criterion AI’s iPhone and Android apps.

Offline deployments require a suitable on-premise environment, which typically consists of some GPU-powered hardware installed close to the place of operation (e.g., right next to the production line). Our models are exported to TensorFlow’s SavedModel format, which is an industry standard for serializing models and enables the use of TensorFlow Serving.

Learn more about how to deploy models in Criterion AI →

Integration with other IT systems

Most IT environments of pharmaceutical companies are fairly complex. For that very reason, we have made the effort to integrate with the most common IT systems out there, including typical Microsoft-based on-premise environments as well as public cloud service providers such as Amazon Web Services, Microsoft Azure and Google Cloud Platform.

In addition, we have a partnership with Bigfinite, a company specialized in collecting, storing and activating data from production lines with the purpose of using them to train AI models. Data collected by Bigfinite can easily be imported into Criterion AI and models trained can subsequently be exported back to Bigfinite.

Learn more about how to easily integrate Criterion AI with other systems →

?
This website collects usage data to help us improve your experience of the site. Click here to accept or here to opt out.
×