The paper describes the prototype of a novel embedded 3-axis force sensor which is intended to be used for detecting and measuring the contact forces at the fingertips of a robotic hand when grasping and manipulating objects. The sensor is composed of three main parts: a printed circuit board with Hall effect sensors, a neodymium magnet and a elastic silicon layer. The dimensions of the sensor that should be placed at the fingertip are minimized to fit the size of a human inspired robotic hand. The signal processing of the data obtained from the Hall effect sensors is completely done with an ARM Cortex-M4 micro-controller with implemented neural network. The target data which is used for training the neural network is obtained from reference precise 6 axis force/torque sensor. The experimental setup as well as the procedure for acquiring the training data set, learning and implementing the neural network on embedded platform are presented.