Multimodal Physics-Informed Foundation Model for Air Quality

University of Surrey; Guildford; On-site; GBP 45,000 - 60,000; Part time

Job summary

University of Surrey invites applications for a world-class AI researcher to lead the development of a physics-informed, multimodal Foundation Model for air quality. You will work at the intersection of deep learning and atmospheric science, collaborating with Surrey's PAI, CVSSP and GCARE teams.

You will design, pre-train, and fine-tune large models on HPC resources, embedding physical constraints to ensure interpretability and policy relevance, while delivering web-based proof-of-concept tools

Qualifications

• Experience in designing and training large-scale DL architectures (e.g., Transformers, similar spatiotemporal models).
• Proven experience in multimodal data fusion (time-series, imagery, tabular).
• Strong proficiency in PyTorch and TensorFlow; experience deploying models on HPC clusters.
• Experience building web applications to operationalise models for policy decisions.
• Solid mathematical foundation in calculus and linear algebra; implement physics-informed constraints in NN losses.
• Track record of publishing high-quality research in AI or environmental science venues.

Responsibilities

• Architect a future-facing FM for air quality prediction fusing satellite, surface sensors, meteorological data.
• Embed physics constraints into model architecture for interpretability and trust.
• Apply super-resolution to translate coarse data into street-level insights.
• Fine-tune the pretrained FM for downstream applications and web-based proof-of-concept.
• Pretrain and fine-tune on Surrey's HPC infrastructure.
• Collaborate with national and international partners across academia and industry.

Skills

DL architectures

Multimodal fusion

PyTorch

TensorFlow

HPC deployment

Web apps

Math foundations

Research publishing

Physics-informed constraints

Tools

NVIDIA GPUs

Job description

University of Surrey invites applications for a world-class AI researcher to lead the development of a physics-informed, multimodal Foundation Model for air quality. You will work at the intersection of deep learning and atmospheric science, collaborating with Surrey's PAI, CVSSP and GCARE teams.

You will design, pre-train, and fine-tune large models on HPC resources, embedding physical constraints to ensure interpretability and policy relevance, while delivering web-based proof-of-concept tools