Hemodialysis is one of the treatment methods for the patients with end stage renal disease. It is conducted through the use of artificial kidney or dialyzer, located outside of the human body. During the hemodialysis therapy, small air bubbles may infiltrate through the blood tubing and coalesce to perform the larger bubble which can be harmfull for the patient if entering the patient’s blood circulatory system. The objective of this work is to develop an ultrasonic based air bubble detection system as well as the safety system, mainly the electronic system, for hemodialysis machine application. The safety systems consits an electronic for activating visual and sound alarms, and tube clamp. It is used for preventing the embolism during the hemodialysis therapy. The research method covers the conceptual design, detail design, prototyping and performance testing for both hardware and software. The hardware system consists of air detector module, analog circuitry which consists of transmitter and receiver, digital circuitry with DSP based SOM (system on module) and I/O interface. The software system has two main functions. It is for controlling the hardware and processing the signals digitally. The integration between the hardware and the software making the system be able to detect the presence of air bubble. In addition, the software had possess the capability for triggering a tube clamp as well as activating visual and audio alarms. The tubing clamp was used to stop the fluid flow in the blood tubing and prevent an air bubble from entering the patient circulatory system. The validation test was conducted to verify the functionality of the system. For testing purposes, an experimental flow loop was constructed. This loop consits of peristaltic pump, bubble injector, tube clamp, air flow meter, liquid flow meter, and bubble detector. Several test runs by using simulated blood (Dextran 70 solution) were performed. To test the performance of bubble detection system, comparison between measurement of PWM signals conducted using the DSP and oscilloscope was presented in this work. It shows that error procentage between two types of measurement is less than 5%.