AI for Science Research Engineer

NATIONAL UNIVERSITY OF SINGAPORE; Singapore; On-site; SGD 120,000 - 180,000; Full time

Job summary

The National University of Singapore seeks a skilled ML systems engineer to design the AI for Science Gym substrate. You will bridge scientific data, ML representations, and compute environments used daily by researchers, shaping scalable, reusable workflows on HPC.

Responsibilities include building data pipelines, modular model interfaces, and tokenization pipelines, with a focus on end-to-end reproducible ML workflows and active collaboration with domain scientists.

Qualifications

• Degree in a Computational, Engineering, or Physical-Science discipline; a Master's or Bachelor's with equivalent research-engineering experience.
• Experience building data pipelines and reproducible ML workflows end to end, not just training scripts.
• Comfort with GPU compute and scientific data formats (HDF5, TIFF/OME-TIFF, Zarr).
• Track record of modular, maintainable software — well-tested and well-documented.

Responsibilities

• Design and implement data pipelines for scientific datasets (electron microscopy, X-ray, light microscopy).
• Develop modular ML model interfaces and tokenizers for interchangeable components.
• Create baseline tokenization and cartography pipelines as reference implementations.
• Build LLM- and RAG-assisted interfaces for dataset navigation and workflow discovery.
• Ensure forward-compatibility to HPC with containerised, scalable deployment (Slurm, Docker/Apptainer).
• Onboard and train new Gym users; help instructors build reusable workflows.

Skills

Python

PyTorch

ML workflows

LLMs & RAG

HPC deployment

System design

Education

Master's degree

Bachelor's degree with equivalent research-engineering experience

Tools

Docker

Apptainer

Slurm

HDF5

Job description

The National University of Singapore seeks a skilled ML systems engineer to design the AI for Science Gym substrate. You will bridge scientific data, ML representations, and compute environments used daily by researchers, shaping scalable, reusable workflows on HPC.

Responsibilities include building data pipelines, modular model interfaces, and tokenization pipelines, with a focus on end-to-end reproducible ML workflows and active collaboration with domain scientists.