About the team 

Predictive AI and Data team is responsible for providing AI and Bioinformatics solutions to the scientists across the spectrum of drug development and discovery in AstraZeneca (both pre-clinical and clinical stages). The primary aim is to find ways to accelerate the drug development process by leveraging existing company data in combination with the most cutting-edge AI approaches in day-to-day scientific work across the company. 

Introduction to role 

The Associate Director will apply cutting edge advanced computational pathology and AI technologies to multimodal digital pathology datasets and help derive biological insights from combination of multimodal model inference from pathology, omics (transcriptomics, cell paintint and proteomics) and chemical/ADME foundation models for drug safety prediction to support drug portfolio project milestone decision making.  

Accountabilities 

• Support Computational Pathology foundation model development: data preparation including stain/scan harmonization & standardization, data augmentation, rigorous QC, WSI tiling/patch extraction.   

• Build and evaluate multi-resolution architectures (hierarchical, pyramid, or attention-based models) that integrate tile and slide-level context across 5×–40× magnifications for robust morphological feature learning.   

• Undertake self-supervised representation learning: Train foundation models on multi-scale WSIs using scalable distributed training and modern practices tailored to histopathology.   

• Implement, where appropriate, weak/multiple-instance learning (MIL): Implement and optimize MIL and weakly supervised strategies for downstream tasks where only slide- or case-level labels are available.   

• Operate multi-GPU/cloud enviroments ensuring reproducibility at scale: Distributed learning across GPUs, containers, orchestration, feature stores, versioning, experiment tracking, and checkpointing. 

• Evaluation benchmarking for representation quality (linear probe, few-shot, retrieval), cross-site generalization, stain/scan robustness, and uncertainty calibration with statistically sound comparisons.   

• Integrate multimodal representations: Investigate approaches to integrate/fuse pathology, Cell Painting, cell-assay omics and chemical/ADME features e.g. via joint embeddings, late fusion, and cross-modal attention. 

• Support development towards unified drug safety risk scores using ensembling/meta-learning, modality ablations, OOD detection, and prospective validation on preclinical toxicity datasets. 

• Collaborate closely with pathologists, computational safety AI scientists and other safety SMEs; drive interpretability reviews, clear communication to leadership, and milestone planning. 

• Scientific communication: Drive high-quality and open science & logistics communication across project team to ensure whole team is aligned throughout project 

Essential Skills/Experience 

• PhD in Computer Vision & AI including computational pathology 

• Strong hands-on ML/DL experience (preferably using PyTorch). 

• Experience with self-supervised learning/MIL model training approaches 

• Advanced Python programming skills. 

• Experience with pathology Whole Slide Image (WSI) datasets and preparation & use for model training 

• Proven delivery of computational pathology model impact. 

• Excellent communication and stakeholder engagement skills. 

• Solves complex problems using scientific judgement. 

• Builds strong cross-functional relationships and networks. 

Desirable Skills/Experience 

• Experience with Omics, Cell Painting or other biomedical imaging data. 

• Background knowledge or experience working with pathology or similar medical imaging datasets. 

• Experience with cloud computing and workflow automation.