How to generate synthetic data using MOSTLY AI’s Platform?

It all starts with the creation of a Generator. A Generator needs original (e.g. production) data to learn from. This original data as an input can either be uploaded as a file or configured to be read from a Connector (Data Source). Data is provided in tables and you can add as many tables as you wish.

If the tables are connected to each other you can easily configure primary and foreign keys and relationships. As the last step you need to configure the Models that will be trained. But don't worry - the standard settings will most often be just fine and maybe you only want to define the "Maximum training time". That's the most the model training will last. You can also just use one of the Configuration presets: Accuracy, Speed, Turbo.

Once you've completed the configuration of the Generator a click on "Start training" will kick off the process of creating a Generator. The Platform will run through a couple of steps incl. the training of the Generative AI model for every table that you can follow in the Training status section. This is all done automatically and you just have to sit back, relax, and wait. But don't get yourself too comfortable because of the high performance of the Platform you're Generator will be ready in no time!

You can then review the Model Insights report to see if you're happy with the results. If you are, your newly created Generator is ready to be shared with the world. Just click on "Share", enter the email address of the person who you want to share the Generator with and set the permission level. And voilà - the Generator will appear in the account of the person ready to be used.

A "Data Consumer" will use such a Generator to finally create a Synthetic dataset. First a couple of things need to be configured (e.g. how much synthetic data should be created) to make sure the output is exactly as wished. Then it's as simple as clicking "Start generation". The Platform will create the desired synthetic data fully automatically. Once this is complete the synthetic data can be explored via the Data Report, downloaded or stored in a target system for further use.

Now the magic doesn’t stop here. Quite the contrary. Once you have your synthetic data created (and even earlier, if you only have a Generator trained) you can start exploring using the Assistant. Just click on “Explore” and use a natural language interface to ask questions about your data, perform analyses, create charts and visualizations, and to get some tangible insights. All with the peace of mind that the synthetic data is fully anonymous and the privacy of any individuals fully protected.