Research Fellow (Humanities & Social Studies Education Department) [NIE]

The National Institute of Education invites suitable applications for the position of Research Fellow on a 25-month contract at the Humanities & Social Studies Education Department.

Project Title: FastTrackAI: Fast-Tracking AI for Scientific Discovery with Accelerated Networking and Computation on Modern Architectures in Singapore

Project Introduction: The advancement of accelerated computation and communication technologies is driving the progression of modern AI. However, current efforts are predominantly concentrated on Large Language Models (LLMs) tailored for industrial applications, leaving AI for Science with inadequate attention and support. This proposal addresses the gap by focusing on the development of optimized implementations for AI for Science workloads. Bespoke communication and computation libraries specific to particular architectures are often difficult-to-use as soon as one deviates from mainstream language models to multi-model AI for Science workloads. The result can be inefficient and slow training and inference, increasing costs and decreasing researcher productivity by orders of magnitude. To overcome these challenges, we propose to design and develop the first performance-portable software optimization and communication stack for AI for Science workloads. The compute acceleration is achieved using dataflow optimization techniques that integrate with scientific simulations to enable efficient use of new capabilities such as combination of simulation and AI models. The communication acceleration will be compatible with both current and next-generation communication infrastructures, such as InfiniBand and Ultra Ethernet. Our project will also deliver a comprehensive set of PyTorch libraries, encompassing various optimized models for scientific applications, including Graph Neural Networks (GNNs), Convolutional Neural Networks (CNNs), and Transformers. These resources will be made available to researchers across different scientific domains. Ultimately, this initiative will empower scientists to effectively leverage the computational infrastructure available in Singapore, addressing the most complex and demanding AI for Science problems with enhanced efficiency and efficacy. With Singapore pledging to spend one billion dollars on AI in the near future and possible performance optimizations of more than 10x, the impact of our work could be transformational.

Requirements:

• PhD in Hydrology, Computing, Earth science or a related field

• Proficiency in remote sensing, geographic information systems, and statistics

• Proficiency in programming with Python, R, and Google Earth Engine

• Minimum 3 years of relevant working experience in an academic environment

• Experience in academic report writing or paper publication

• Experience with geography fieldwork

• Able to work independently and possess strong follow-through skills

• A team player in a multi-cultural environment

• Strong communication and organization skills

• Strong analytical, problem-solving and writing skills
  

Responsibilities:

• Conduct independent and collaborative research work

• Manage research project

• Liaise and work with research stakeholders and vendors

• Prepare literature reviews to synthesize prior studies

• Propose, design, and implement appropriate research methodologies

• Remote sensing and field data collection, storage and archival

• General quantitative data analysis

• Writing of reports, publications, and presentation materials

• Planning and executing tasks to achieve key deliverables

• Any other administrative duties as assigned by Principal Investigators
  

Application

Applicants (external and internal) will apply via Workday. We regret that only shortlisted candidates will be notified.
Workday Job Requisition ID: R00025638

(https://ntu.wd3.myworkdayjobs.com/Careers/job/NTU-Main-Campus-Singapore/Research-Fellow--Humanities---Social-Studies-Education-Department---NIE-_R00025638)

Closing Date

Closing date for advertisements will be set to 14 calendar days from date of posting.