nitroproc allows you to do ML, AI and data science on your mobile device using massive datasets. You can pre-process your data, or use techniques such as random forests or gradient boosting. Because it uses the disk, you won't typically have memory issues, and will be able to work with millions of samples. Use the included examples for doing classification, regression, and data preparation. Apart from those, you can upload new datasets to your device using a USB cable (place your files on /Downloads).
You can either code your scripts (using the editor) or use the integrated GUI. The latter option will certainly be much easier on touch-screen devices.
nitroproc also runs on PC/Macs so you can deploy the same scripts on those machines (naturally your code will run faster there)
Currently, it only accepts csv files and produces csv files as outputs. No internet connection is required, and no server is used - everything runs locally on your phone or tablet. Your data is safe, as absolutely nothing is uploaded to any server.
The app also contains a very nice navigation system for your scripts and csv files. You can open each data file using the app included viewer.
Every nitroproc script generates a log file that reports any issues and performance metrics for each instruction, and a logtracer file that describes the logical connection between the instructions you are running (this is useful when you do lots of transformations to the files and you lose track on how the different files are connected).