Open role
Intern, Architecture Research Engineer (MoE Placement for Serving at Scale)
Samsung Semiconductor
About this role
Join Samsung Semiconductor's AGI Computing Lab as an Architecture Research Intern to tackle the complex system-level challenges of future AI/ML workloads. You will research and co-optimize expert placement for Mixture-of-Experts (MoE) models in distributed systems, contributing to the design of scalable AI serving platforms and potentially publishing research papers.
What we are looking for 6
- Research MoE expert placement, replication, and migration for distributed serving
- Co-optimize expert placement with parallelism, memory hierarchy, and interconnect
- Characterize expert load imbalance and temporal drift on real workloads
- Extend analytical simulators and DSE tools to model MoE serving
- Integrate placement-aware scheduling with compiler and runtime teams
- Contribute to ML systems and architecture research, potentially publishing papers