Brain-machine interfaces (BMIs) have emerged as a transformative force in assistive technologies, empowering individuals with motor impairments by enabling device control and facilitating functional recovery. However, the persistent challenge of inter-session variability poses a significant hurdle, requiring time-consuming calibration at every new use. Compounding this issue, the low comfort level of current devices further restricts their usage. To address these challenges, we propose a comprehensive solution that combines a tiny CNN-based Transfer Learning (TL) approach with a comfortable, wearable EEG headband. The novel wearable EEG device features soft dry electrodes placed on the headband and is capable of on-board processing. We acquire multiple sessions of motor-movement EEG data and achieve up to 96% inter-session accuracy using TL, greatly reducing the calibration time and improving usability. By executing the inference on the edge every 100ms, the system is estimated to achieve 30h of battery life. The comfortable BMI setup with tiny CNN and TL paves the way to future on-device continual learning, essential for tackling inter-session variability and improving usability.