Real-time processing of ECG, EEG, and speech signals is very important in portable biomedical applications. AI is widely used to improve the performance of Signal processing applications. Delay, power, and accuracy are the major design constraints in the designs of portable devices for biomedical signal processing applications. Most portable systems acquire ECG/EEG signals and transmit them for further analysis. A real-time low complex and low delay machine learning-based architecture to acquire signal ECG from the heart, process, analyze and predict/detect abnormalities in the signal can save the life of many patients. A real-time and low complex device for accurate Epileptic seizure forecasting based on EEG signals can improve the quality of life for people with epilepsy. Software implementations of machine learning algorithms on a general purpose processor are relatively slow due to sequential and temporal exaction. Hardware implementation offers high performance due to parallel and spatial execution. FPGA ( Field-programmable Gate Arrays) can inherently provide low delay and reconfigurable computing for complex real-time Applications. Microsoft Azure cloud is based on Intel's FPGA devices and offers performance, and flexibility for complex computations. This seminar gives an overview of challenges in machine learning for signal processing applications and high-performance and reconfigurable computing using FPGAs.