In this article, we have presented a wearable device for capturing and identifying hand-gestures as an input modality, which can be used in OS platform independent generic HCI application in multi-modal platform through its customized action-activity mapping. We have used Tiny ML supported micro-controller development board with integrated motion sensor for developing the hardware prototype. The developed prototype has ability to identify and follow a set of hand-gestures, which are convenient to users, for operating OS platform independent machines or computers in a multi-modal input-interface. The design procedure is consisted of design methodology for hand-gesture selection, design methodology for hand-gesture identification and design procedure for action-activity mapping. We have considered six static and four dynamic hand-gestures based on the convenience of the users for the modal operations of the gestures. The static gestures give 100% accuracy in their operations. The dynamic hand-gesture classification employs classical machine learning algorithms. The Random Forest classifier provides a remarkable accuracy in the said applications. The developed prototype has been tested in real application operated in Android OS, Windows OS, and Raspbian OS. These findings show the potential of hand-gesture based input modality for OS platform independent generic HCI application in multi-modal platform.