Physics-Driven AI Research Scientist: PINNs & UQ

WESTERN DIGITAL (SINGAPORE) PTE. LTD.; Singapore; Presencial; SGD 120.000 - 240.000; Jornada completa

Descripción de la vacante

Western Digital (Singapore) Pte. Ltd. seeks a research scientist to design physics-constrained ML for product development.

You will originate PINNs, uncertainty quantification, and causal ML methods, validating with real lab data. You will work with ML engineers to shape what is built, publish strong research outputs, and produce complete technical documentation for handoff to product teams. PhD preferred; mastery in PyTorch/JAX; publications or open-source contributions valued; Singapore-based

Formación

• Master's or PhD in AI/ML, Physics, Applied Math, or related field.
• 1–3 years research experience for Master’s; PhD open to all.
• Peer‑reviewed publication or significant open‑source scientific ML contribution.
• Strong collaboration and documentation capabilities.

Responsabilidades

• Originate PINNs methodology and design physics‑constrained loss architectures.
• Lead Bayesian deep learning, active learning, and ensemble uncertainty work.
• Own causal ML framework development for reliability and yield analysis.
• Design synthetic data approaches (diffusion/VAEs) for validation.

Conocimientos

PyTorch

JAX

PINNs

Bayesian deep learning

Active learning

Structural causal models

Demonstrated research output

Prototype-to-Documentation handoff

Educación

Master's degree or PhD in AI/ML/Physics/Applied Mathematics

Descripción del empleo

Western Digital (Singapore) Pte. Ltd. seeks a research scientist to design physics-constrained ML for product development.

You will originate PINNs, uncertainty quantification, and causal ML methods, validating with real lab data. You will work with ML engineers to shape what is built, publish strong research outputs, and produce complete technical documentation for handoff to product teams. PhD preferred; mastery in PyTorch/JAX; publications or open-source contributions valued; Singapore-based