Bio:

Thomas Van Vaerenbergh received the master's degree in applied physics and the Ph.D. degree in photonics from Ghent University, Ghent, Belgium, in 2010 and 2014, respectively. He was awarded the scientific prize Alcatel-Lucent Bell/FWO for his PhD thesis on all-optical spiking neurons in silicon photonics. In 2014, he joined the Large-Scale Integrated Photonics Lab in Hewlett Packard Labs, part of Hewlett Packard Enterprise (HPE), in Palo Alto, California.  Since 2019, he is based in HPE Belgium and uses his large network of academic and industrial partners to expand HPE’s research activities related to photonics and AI in the EMEA region (including as HPE’s research lead for the Horizon Europe project NEUROPULS, where CEA LETI is also a research partner). His main research interests include analog photonic and electronic accelerators for combinatorial optimization and AI workloads, and inverse design of photonic devices and circuits based on physics-informed machine learning.

Abstract:

Whenever there is a need for high bandwidth, low latency and energy-efficiency in communication channels, the multiplexing capabilities of optical interconnects offer tremendous advantages over their electronic counterparts. By combining this fundamental advantage of photons over electrons with the CMOS fabrication capabilities developed by the electronics industry, silicon photonics allows such optical interconnects to be sufficiently cost-efficient to become viable for exascale and post-exascale interconnects in High Performance Computing (HPC) systems. To reach more than 1Tb/s in a single fiber using less than a pJ/bit for realistic bit error rates, it is however critical to have on-chip lasers. Therefore, over the last years, HPE has developed a heterogeneous IIIV-on-Si platform, which incorporates all required modulators, filters, and comb lasers (required for wavelength division multiplexing) for such transceiver circuits in the same fabrication flow.

In this talk, I will explain how HPE leverages multiplexing in this heterogeneous IIIV-on-Si platform to develop energy-efficient and high-bandwidth interconnects as well as neuromorphic accelerators targeting AI workloads