A wrist-worn PPG sensor coupled with a lightweightalgorithm can run on a MCU to enable non-invasive and comfortable monitoring, but ensuring robust PPG-based heart-ratemonitoring in the presence of motion artifacts is still an openchallenge. Recent state-of-the-art algorithms combine PPG andinertial signals to mitigate the effect of motion artifacts. However,these approaches suffer from limited generality. Moreover, theirdeployment on MCU-based edge nodes has not been investigated.In this work, we tackle both the aforementioned problems byproposing the use of hardware-friendly Temporal ConvolutionalNetworks (TCN) for PPG-based heart estimation. Starting from asingle “seed” TCN, we leverage an automatic Neural ArchitectureSearch (NAS) approach to derive a rich family of models. Amongthem, we obtain a TCN that outperforms the previous state-of-theart on the largest PPG dataset available (PPGDalia), achieving aMean Absolute Error (MAE) of just 3.84 Beats Per Minute (BPM).Furthermore, we tested also a set of smaller yet still accurate (MAEof 5.64 - 6.29 BPM) networks that can be deployed on a commercialMCU (STM32L4) which require as few as 5k parameters and reacha latency of 17.1 ms consuming just 0.21 mJ per inference.Neural network pruning reduces network complexity andstorage by removing unimportant connections in the network, enabling network miniaturization, fast training andinference, easy deployment to portable devices, etc. The emerging lottery ticket hypotheses and sparse initializationtechnique have shed new lights on the pruning research.However, few research focuses on the pruning of the networks for remote photoplethysmography (rPPG) pulse signal extraction. Opposite to the existing pruning researchesthat prune large network, rPPG networks are relatively small.