Serving Your Model
Once you have developed a model with AI Zwei, you may be interested in looking at predictions made on new data you collect over time; this can help you determine if the model is satisfactory for you needs, and it can help you scrutinize decisions made by the model for specific examples to help you uncover model bias.  It is easy to serve your model with AI Zwei.  After creating a model, or choosing your favorite model from a group, you can ask AI Zwei to serve your model.  After prompted for this, AI Zwei will share a link to Google Sheets with you. 
 
1. Simply click on the link.  Two Google Sheets will appear, one sheet named "Data" and the other "Configs".  On the "Data" sheet, you can enter the features of any new examples, or specific examples, for which you would like to know your model's predicition. Add an empty column to the end (right-most) named as per your convenience where AI Zwei can put the predictions.
DataSheet.png

2. On the "Configs" sheet you will enter your user id and the name of the model from which you would like to elicit predictions.

ConfigsSheet.png

3. Once you have added your data to the “Data” sheet and the required details on the “Configs” sheet, all you need to do is click on the “AI ZWEI” drop-down menu at the top and then click “Run”. This will forward the data to Ai Zwei which will churn the data to get predictions and return them to you in the “Data” sheet right in front of you.

4. Don't worry. If you feel you need assistance, you can go to “AI Zwei” menu and click on “Help”.  This will point you to relevant documentation that will assist you in using this model serving/prediction process.

5.  And if you encounter errors . . . consider this:

If you see “NA” as predictions, most likely there is an error in the data put in the “Data” sheet; for example, perhaps you entered a string in a column that is expected to have numbers, or you forgot to put data in one of the cells. Not to worry, you can edit the data on the fly and click on “Run” again if you feel there are no more data errors.