Doctoral Fellowship in Medical Vision-Language Models

About Internship

Assistant Professor from Indian Institute of Technology, Jammu (IIT Jammu), is accepting interns interested in the field of Computer Science Engineering, in the research area of Artificial intelligence and Machine learning.

This doctoral position is part of a fully funded ANRF (Anusandhan National Research Foundation) research project titled "Mitigating Memorization in Medical Vision–Language Models." Medical Vision-Language Models (VLMs) combine complex clinical imagery (like X-rays or MRIs) with textual reports to assist in automated diagnosis. However, these models often suffer from "memorization"—accidentally storing sensitive patient data, private clinical notes, or exact training images within their neural weights. This position focuses on pioneering novel algorithmic techniques to mitigate data memorization, ensuring that next-generation medical VLMs are private, robust, and safe for clinical deployment. Roles & Responsibilities: 1. Investigate and develop advanced machine learning methodologies (e.g., differential privacy, machine unlearning, contrastive regularization) to prevent memorization in generative vision-language models. 2. Work extensively with large-scale multimodal models, optimizing the interaction between deep vision encoders (like ViT) and language models (LLMs). 3. Build and execute rigorous testing pipelines to evaluate data leakage, membership inference attacks, and privacy vulnerabilities in medical AI systems. 4. Manage and process complex, high-dimensional multimodal medical datasets (e.g., MIMIC-CXR) while adhering to strict privacy and ethical protocols. 5. Author and publish high-impact research papers in top-tier AI conferences (CVPR, NeurIPS, ICCV, ICLR) and leading medical informatics journals. 6. Assist in drafting project progress reports, technical documentation, and milestone presentations for the funding agency.