Research Fellow: AI Systems & Large-Model Optimization

National University of Singapore; Singapore; On-site; SGD 70,000 - 110,000; Full time

Job summary

The National University of Singapore invites applications for a Research Fellow to work on large-model optimization and software-hardware co-design of AI computing systems. The role explores how models are trained and run efficiently, balancing computation, memory, and communication with hardware considerations.

Fellow will develop methods to reduce resource use while preserving model quality, and will collaborate with colleagues across ML, computer systems, and architecture to improve

Qualifications

• PhD required in CS, ECE or closely related field.
• Experience in efficient deep learning, ML systems, compilers, parallel and distributed computing, or computer architecture is preferred.
• Strong programming abilities and ability to turn ideas into working implementations and experiments.

Responsibilities

• Conduct original research in large-model optimization and software-hardware co-design of AI computing systems.
• Identify bottlenecks in training and inference, develop methods, and assess trade-offs between model quality and efficiency.
• Build prototypes and evaluate methods on computing platforms or architectural simulations.
• Publish results and collaborate with ML, systems, and computer architecture teams.

Skills

Programming

Independent research

Communication of findings

ML/AI systems experience

Education

PhD in Computer Science or Electrical and Computer Engineering

Job description

The National University of Singapore invites applications for a Research Fellow to work on large-model optimization and software-hardware co-design of AI computing systems. The role explores how models are trained and run efficiently, balancing computation, memory, and communication with hardware considerations.

Fellow will develop methods to reduce resource use while preserving model quality, and will collaborate with colleagues across ML, computer systems, and architecture to improve