PhD Student: Privacy Enhancing Technologies for LLMs and AI agents(m/w/x)

Description

You will drive innovation in AI security by developing novel privacy methods for LLMs. Your day-to-day involves researching attack vectors and publishing findings to protect data systems.

Requirements

• •Master’s degree in computer science, mathematics, engineering, or related field
• •Background in machine learning, deep learning, or natural language processing
• •Proficient Python skills and deep learning framework experience
• •Familiarity with large language models and agent-based systems
• •Interest in research and rigorous problem formulation
• •Knowledge of ML privacy and security techniques (advantageous)
• •Prior research experience or publications (advantageous)
• •Excellent written and spoken English

Tasks

• •Review state-of-the-art privacy for LLMs
• •Monitor emerging AI agent attack vectors
• •Develop novel privacy methods for agent-based systems
• •Address risks like sensitive data leakage
• •Mitigate prompt-based extraction and unintended memorization
• •Implement proof-of-concept prototypes for proposed methods
• •Evaluate prototypes on public and industrial datasets
• •Publish findings in leading scientific journals
• •Collaborate on interdisciplinary machine learning topics

Tools & Technologies

Benefits