Research Fellow – AI‑Enabled Physics Data Science

About the role

The AI for Science Gym creates bottom‑up AI capability across NUS science and engineering. As a Discovery Gym Lead you act as an AI‑enabled data scientist embedded in a physics department, running weekly peer‑learning sessions for post‑graduates and post‑docs and partnering with research groups to turn instrument data into working AI/ML pipelines and teaching material.

Key responsibilities

• Facilitate weekly small‑group peer‑learning cohorts, guiding participants through real data challenges.
• Onboard new departments as they join the Gym.
• Lead one‑month Discovery Sprints: assess dataset suitability, deliver an interactive dashboard in week 1, define research goals, and hand over scripts and pedagogical datasets.
• Deposit partner data as tokenised pedagogical challenges and contribute domain‑workflow demos.
• Publish sprint outcomes as preprints and compete in AI meta‑harness scrimmages.
• Participate in monthly DGL methods‑exchange and weekly syncs with the Architect and other leads.
• Maintain a workload of 1–2 active sprints, with consultation limited to ≤40% of weekly time, and produce at least one paper or proceeding per year.

Required profile

• Ph.D. in a science or engineering discipline (e.g., Physics, Chemistry, Materials Science).
• Post‑doctoral or industry experience applying machine learning in research.
• Strong publication record and a desire to continue publishing.
• Excellent communication and teaching instincts.
• Ability to ship functional dashboards quickly and evaluate dataset quality.

Required skills

• End‑to‑end data‑science and ML pipelines (wrangling, dimensionality reduction, clustering, labeling, supervised and unsupervised learning).
• Scientific Python with modern ML stack: PyTorch or JAX, scikit‑learn, pandas, numpy.
• Version‑controlled, reproducible workflows (Git).
• Experience handling messy real‑world instrument data.
• Familiarity with HPC or multi‑GPU environments.