Innovation Postdoctoral Fellow, Data Science & AI Vision Foundation Models for Cellular Imaging in AI Drug Discovery

Novartis • San Diego, CA

2d • Onsite

About The Position

The Novartis Biomedical Research Postdoctoral Fellowship Program offers a unique training opportunity for exceptional early-career scientists eager to advance AI-powered phenotypic drug discovery. As a Postdoctoral Research Fellow, you will join the Discovery Sciences (DSc) group in San Diego to develop foundation models for high-content cellular imaging, working alongside leading scientists in a highly collaborative, multidisciplinary environment while gaining exposure to the broader ecosystem that translates scientific discovery into medicines. This is a full-time training position of up to three years, starting October 1, 2026. The fellowship aims to advance representation learning for high-content cellular imaging to accelerate AI-powered phenotypic drug discovery across Novartis. The project will develop, benchmark, and deploy foundation models that generalize across assays, perturbations, cell types, and imaging modalities. You will curate large-scale microscopy datasets, pretrain and adapt vision foundation models, and design rigorous evaluation protocols to assess model performance, robustness, and generalization on public and proprietary datasets. You will work closely with cross-functional scientists and engineers to apply these models to answer key scientific questions in the drug discovery pipeline. You will be connected to a large network of data scientists and postdoctoral researchers across the organization. The project will deliver a unified platform for training, benchmarking, and deploying foundation models for high-content cellular imaging. The platform will support critical downstream applications and raise the accuracy, scalability, and reproducibility of phenotypic analysis across Novartis. Methods and results will be published in leading machine learning and biomedical venues.

Requirements

• PhD (or equivalent doctoral degree) in a relevant scientific discipline completed prior to the fellowship start date.
• The program is intended for scientists immediately following their PhD training (PhD conferred in 2026 only).
• Demonstrated record of scientific achievement (publications, presentations, patents, or equivalent)
• Strong commitment to learning, innovation, and professional development
• Ability to formulate and drive independent research questions
• Proficiency in Python and modern deep learning frameworks (e.g., PyTorch, TensorFlow, or JAX), with hands-on experience training models on GPUs.
• Experience training deep learning models on HPC or cloud infrastructure with reproducible code
• Working knowledge of vision foundation models, including self-supervised learning and vision transformers, with experience fine-tuning or adapting pretrained models to new tasks.
• Experience designing benchmarking and evaluation protocols for model generalization.
• Strong written and verbal communication skills

Nice To Haves

• Experience pretraining vision foundation models from scratch on large-scale imaging datasets.
• Expertise in high-content or biomedical imaging (e.g., Cell Painting, phenomics, digital pathology) and channel-adaptive or multimodal architectures.
• Experience with distributed training on HPC clusters, reproducible ML pipelines on cloud infrastructure, and model-tracking tools such as MLflow
• Publications at top ML venues (NeurIPS, ICML, ICLR, CVPR) or biomedical ML venues (MICCAI, etc.)

Responsibilities

• Develop vision foundation models for high-content cellular imaging using state-of-the-art self-supervised learning techniques.
• Design robust evaluation protocols that measure how well models generalize to unseen assays, perturbations, cell types, and imaging conditions.
• Curate large-scale public and internal microscopy datasets and establish standards for data quality and reproducibility.
• Extend the models to multimodal settings by integrating imaging with other data modalities, such as compound data, for drug discovery applications, such as phenotypic profiling.
• Build scalable training and inference pipelines across high-performance computing and cloud infrastructure with clean, reproducible, and well-documented code.
• Propose innovative modeling and evaluation methods that advance the team's technical direction.
• Work independently and collaboratively with data scientists across disease area, platform, and AI research teams, and publish and present results.

Benefits

• health, life and disability benefits
• a 401(k) with company contribution and match
• a variety of other benefits
• generous time off package including vacation, personal days, holidays and other leaves

What This Job Offers

• Job Type: Full-time
• Career Level: Entry Level
• Education Level: Ph.D. or professional degree