PhD defence by Yevhenii Osadchuk

Abstract

The rapid growth of cloud services and large-scale AI models is driving demand for high-speed, energyefficient data center communications. Intra-datacenter optical fiber systems provide a cost-effective solution, but the noise mitigation block in the receiver presents a challenge due to its high computational complexity. ML-based equalizers outperform traditional methods, offering improved performance with reduced complexity. Low complexity ML-based techniques are developed, providing scalable, energy-efficient noise mitigation techniques suitable for short-reach data center connectivity in the upcoming 800G era. Reduced energy consumption within data center communication links.

Supervisors

• Main Supervisor: Associate Professor Francesco Da Ros, DTU Electro.
• Co-supervisor: Professor Darko Zibar, DTU Electro.

Assessment committee

• Associate Professor Michael Galili, DTU Electro (chair).
• Professor Adonis Bogris, University of West Attica, Greece.
• Dr. Xi (Vivian) Chen, Nokia Bell Labs, USA.

Master of the Ceremony

• Senior Researcher Deming Kong, DTU Electro.