Research fellow in optical switching for AI data centres

Post Status: Specific Purpose Contract – Full time 

Research Group / Department / School: School of Computer Science and Statistics, IRIS Network AI and Sensing Research Group, Trinity College Dublin, the University of Dublin 

Location: Trinity College Dublin, School of Computer Science & Statistics, Trinity College Dublin, the University of Dublin College Green, Dublin 2, Ireland 

Reports to: Prof Marco Ruffini 

Salary: Appointment will be made on the, SFI Team member salary scale, new postdoctoral researcher, level 2A or Experienced Post Doctoral Researcher (Level 2B), depending on experience, at a point in line with Government Pay Policy [€51,176 - € 55,399 per annum], appointment will be made no higher than point 1, level 2B. 

Hours of Work: 39 hours/week 

Closing Date: 12 Noon (Irish Standard Time), 1st June 2026

Post Summary 

The exponential growth of AI and high-performance computing workloads is placing unprecedented demands on data centre interconnect (DCI) infrastructure, with interconnect power consumption now exceeding that of compute in many deployments. Optical switching has emerged as a transformative approach to address this scaling challenge, but realising its potential requires moving beyond fixed network configurations toward architectures that can dynamically adapt to the traffic patterns of the applications they serve. 

AI and HPC workloads exhibit strong structural regularity in their communication patterns: collective operations, parameter synchronisation, and pipeline stages produce traffic flows that are far from random. This predictability opens an opportunity: by characterising traffic patterns in real time and reconfiguring the optical topology to match them, the interconnect can be tuned to the workload rather than provisioned for the worst case. Realising this requires intelligent network control capable of building dynamic models from live traffic, predicting near-term demand, and translating predictions into wavelength tuning, optical path reconfiguration, and hybrid optical/electronic switching decisions on millisecond-t-omicrosecond timescales. 

This research will develop machine learning models for traffic pattern recognition and prediction in AI/HPC data centres, and integrate them into a multi-layer control plane that drives optical topology reconfiguration. The work will use real workload traces from industry partners and target measurable improvements in GPU utilisation, end-to-end latency, and energy per bit, contributing to the broader goal of an order-of-magnitude reduction in DCI power consumption. 

One Post Doctoral researcher position is available to work in the area of optical switching for AI data centres. 

The applicants should have a solid theoretical background on machine learning, optical networks and network switching and be willing to engage in testbed-oriented research. The work will be carried out in the OpenIreland research infrastructure, the large-scale testbed for experimentation in open optical, wireless and cloud technologies, which is part of the IRIS Network AI and Sensing Research Group. 

The position will be based in the School of Computer Science and Statistics, with Prof. Ruffini research team and affiliated with the ADAPT research centre on AI, at Trinity College Dublin, Ireland. The researcher will have the opportunity to build their research portfolio within a range of projects within the team. 

The position will be under the direction of Prof. Marco Ruffini

Person Specification 

The appointable candidate will hold a PhD, preferably in a Computer Science or Electronic Engineering or related discipline (preferably with a few months of postdoctoral experience). 

Qualifications 

PhD or equivalent experience. The post is applicable to both new or experienced PhD holders, and salary will be commensurate with experience and achievement. The successful candidate will join a team of highly skilled and innovative researchers in next generation optical and wireless networks.