The explosive growth in data-centric artificial intelligence related applications requires a radical shift from traditional von Neumann computing architectures, based on separate processing and memory units. Performing vector-matrix multiplications (VMMs), the most energy-expensive and recurrent operation in AI’s tasks between physically separated processors and memories, results in power and performance bottlenecks. This demands a tremendous effort in establishing more energy-efficient AI hardware, such as neuro-inspired chips that mimic the behavior and the efficiency of biological neurons and synapses. Several device concepts and technologies relying on diverse physical mechanisms and materials are under study for the implementation of neuromorphic hardware. Phase-Change Memory (PCM) based on GeTe-Sb2Te3 alloys and filamentary Resistive RAM (ReRAM) are the two most promising candidates to emulate artificial synapses . In ReRAM, the mapped synaptic weight conductance depends on the rupture/formation of oxygen vacancies conductive paths, while in PCM, the conductance relies on an amorphous-to-crystalline phase transition in the material. Another interesting phase-change material is vanadium dioxide VO2, which has gained interest for the realization of neuro-inspired oscillators operating with remarkable electrical and optical switching properties induced at low power and near room temperature (68 ˚C) [2,3]. However, improvements in the device characteristics remain a necessity for optimal hardware acceleration in both performance precision and energy consumption. Device requirements are also dictated by algorithmic considerations, highlighting the multi-dimensional nature of neuromorphic accelerator designing. These device building blocks can be arranged into different configurations to allow for more efficient computation of heavy AI workloads. The resulting networks of artificial neurons and synapses must fulfill important requirements, such as the scalability and the dense integration in the Back-End-of-the-Line (BEOL) of existing Complementary Metal-Oxide (CMOS) technology, to create a high-performance signal processing accelerator [2,4]. In this talk, I will give an overview on the different nature of the phase transitions in GeTe-Sb2Te3-based alloys and in VO2 films to highlight/discuss how their different physical properties can be exploited to realize artificial synapse- and neuron-like devices, respectively. The material and device engineering will be discussed from different perspectives and put into a larger context of Deep Neural Networks (DNN) and Oscillating Neural Network (ONN) systems for accelerating computing. Keywords: Unconventional Computing, Phase-Change Materials, Resistive RAM, Deep Neural Networks, Oscillating Neural Network, Material & Device Engineering REFERENCES 1. A. Sebastian et al., "Memory devices and applications for in-memory computing", Nature Nanotechnology, 15, 529-544, 2020. 2. From: https://spectrum.ieee.org/analog-ai. 3. L. Posa, G. Molnar, B. Kalas, Z. Baji, Z. Czigany, P. Petrik and J. Volk, "A Rational Fabrication Method for Low Switching-Temperature VO2," nanomaterials, vol. 11, no. 212, 2021. 4. G. Csaba and W. Porod, "Coupled oscillators for computing: A review and perspective," Applied Physics Review, 3 January 2020.