PhD Position in Multi-Level Multi-Objective Design Space Exploration for Chiplet-Based Systems

The Inria Centre at Rennes University is offering a PhD position focused on addressing the design complexity of chiplet-based systems. As Moore's Law slows, chiplet architectures have emerged as a modular solution to improve performance and integration density, though they introduce significant challenges in design space exploration. The successful candidate will develop a formal modeling framework and a multi-fidelity evaluation strategy to optimize power, performance, area, and thermal metrics. This research is conducted in collaboration with CEA under the PEPR ChipMosaic - Phoenix project. The role requires a highly motivated researcher to work within the TARAN team, contributing to state-of-the-art architectural design methodologies.

• Develop a formal modeling framework to capture hierarchical interdependencies in chiplet-based designs.
• Create a multi-fidelity evaluation strategy integrating heterogeneous simulation tools for early-stage exploration.
• Adapt multi-objective optimization algorithms, such as genetic algorithms and reinforcement learning, for chiplet design.
• Validate the proposed framework using representative workloads to demonstrate improvements in power, performance, and thermal metrics.
• Present research progress regularly within the TARAN research team.
• Collaborate with external partners, specifically CEA, under the PEPR ChipMosaic project.