Senior Research Fellow – Kernel Optimization for LLMs

About the role

The School of Computing at the National University of Singapore is seeking a Senior Research Fellow to lead the development of a high‑performance kernel optimisation engine for large‑language‑model (LLM) inference services. Working closely with the Principal Investigator, you will drive research and implementation of automated profiling, bottleneck diagnosis, kernel generation, and performance benchmarking, and integrate the solution into production inference frameworks.

Key responsibilities

• Design and implement a kernel optimisation engine targeting GPU‑accelerated LLM inference.
• Develop automated profiling tools to identify performance bottlenecks.
• Generate and optimise GPU kernels using CUDA and Triton.
• Benchmark and evaluate optimisation results across diverse workloads.
• Collaborate with the research team to integrate the engine into existing inference pipelines.
• Publish high‑impact research findings in top conferences.

Required profile

• PhD in Computer Science or a related discipline.
• Strong background in computer systems and GPU programming.
• Demonstrated research experience with a solid publication record, including first‑author papers at leading conferences.
• Excellent written and spoken English communication skills.

Required skills

• CUDA programming.
• Triton language.
• PyTorch and LLM inference frameworks.
• GPU kernel optimisation and performance benchmarking.

What we offer

• Competitive monthly salary ranging from SGD 7,500 to SGD 9,500.
• Full‑time professional appointment within a world‑class research environment.
• Opportunities to collaborate internationally and contribute to cutting‑edge AI research.