Different combinations of hardware setups and machine learning models were used on a non-perfused breast tissue model to estimate the progress of radiofrequency ablation of tissue based on electrical impedance data. The hardware setup that both implemented and controlled the ablation process through an electrode device was a low-cost embedded system that was designed by our team and allowed for faster and larger data collection, along with reconfigurable predictor support. It consists of the Beaglebone Black (Texas Instruments, Dallas, TX) and an accessory board that contains power switching, impedance, and temperature measurement circuits on-board. The complex electrical impedance measurement subsystem is based on the AD5933 (Analog Devices, Norwood, MA) impedance analyzer integrated circuit, which is designed to measure the impedance magnitude and phase within 2% error range for a multi-frequency range from 10 kHz to 100 kHz, following the low-impedance-ranged CN-0217 reference design from Analog Devices. Using the impedance data from multiple frequencies, different machine learning pipelines were developed based on the idea of data fusion. The problem was posed as a regression task and the ablation depth became the target value. Ultimately, deep neural networks (DNN) and tree-based ensemble (TE) models such as Random Forest and Adaptive Boosting models proved the most useful for the data at hand. The prediction performance showed a mean difference against physical measurements of 0.5 mm for the DNN-based model and 0.7 mm for the TE-based model. Both models were translated to the embedded system, allowing data collection and monitoring the ablation depth by machine learning inference during the ablation process.