PhD in Communication-Optimized Inference in Large Language Models

This PhD position focuses on communication-optimized inference for large language models within the Network Modeling and Design Lab at Eindhoven University of Technology. The successful candidate will combine analytical network modeling using Network Calculus with experimental research on programmable network hardware such as P4 switches and SmartNICs. The project is part of the EuroTech programme and includes a mandatory six-month research stay at the Technical University of Munich. The candidate is expected to publish research findings, release open-source tools, and contribute to energy-efficient AI infrastructure.

• Develop fast and composable analytical network models using Network Calculus to capture inference traffic interactions.
• Investigate how architectural choices like network topology and bandwidth affect performance and energy consumption.
• Design runtime decision tools for workload placement, routing, and resource reservation in AI clusters.
• Conduct reproducible experiments on programmable networking testbeds including P4-programmable switches and SmartNICs.
• Collaborate with external partners such as SURF to evaluate experimental results on real AI infrastructure.
• Contribute to networking standardization communities and release research tools openly.