I like that it's easy to use, with simple drag-and drop features for creating new datasets or adding existing ones from other sources (like Google). The interface isn't as intuitive when you're just getting started; there are many functions buried in menus rather than being obvious at first glance - this can be frustrating if what your looking to do requires some level of customization beyond simply copying/pasting files into an input folder. It would also help tremendously having access to training examples so one could train their models without needing external resources such as AWS instances etc.. Also we had issues where our model was not able properly identify certain text strings within documents even though they were present but weren’t included during creation time via API calls by us nor did exist externally through any file source available online! This has been fixed now however still wish those additional options existed initially instead of requiring manual changes after uploads have.