Postdoctoral Research Associate - Privacy Preserved Federated Learning Algorithms

Requisition Id 16841 

 

Overview:

Oak Ridge National Laboratory is the largest US Department of Energy science and energy Laboratory, conducting basic and applied research to deliver transformative solutions to compelling problems in energy and security.

 

The Discrete Algorithms Group at Oak Ridge National Laboratory (ORNL) seeks a postdoctoral researcher specializing in federated learning and privacy-preservation algorithms. The successful candidate will develop cutting-edge differential privacy techniques for large-scale models across multiple institutions. This position offers a unique opportunity to work with the world's first exascale system, the Frontier supercomputer, and collaborate with experts in machine learning, optimization, electric grid analytics, and image science.

 

The successful candidate will design and implement differential privacy solutions for large-scale scientific data models in federated learning environments. You will advance privacy-preserving machine learning by developing efficient techniques that maintain robust privacy guarantees while minimizing performance impact. Additionally, you will optimize the balance between privacy and utility, addressing the challenges of heterogeneous privacy budgets and varying requirements across diverse clients.

 

Major Duties/Responsibilities:

• Develop and apply differential privacy for large-scale models scientific data to advance research efforts across scientific systems.
• Develop and apply federated learning on distributed and heterogenous datasets.
• Develop more efficient and resilient DP techniques that minimize performance loss while still providing robust privacy guarantees.
• Develop novel privacy-preservation methods that accommodate the diverse privacy requirements of a large number of clients.
• Develop novel mathematically rigorous approaches to optimize the trade-off between privacy and utility especially in the context of large models.
• Advance knowledge of key AI methods such as deep learning, algorithm design, probability theory, privacy definitions, and apply it to develop efficient privacy preserved federated learning model.
• Communicate and coordinate experimental results with other domain experts to facilitate collaboration.
• Present and report research results and publish scientific results in peer-reviewed journals or conferences.

 

Basic Qualifications:

• A PhD in Computer Science, Applied Mathematics, Computational Science, or related discipline.
• Demonstrated hands-on experience and understanding of developing and applying privacy preservation methods to ML models.
• Demonstrated research experience with AI and ML techniques.

 

Preferred Qualifications:

• Knowledge of Approximate, Local, Rényi, Bayesian differential privacy, and other related definitions.
• Knowledge of federated learning SOTA algorithms.
• Knowledge of distributed optimization and consensus algorithms.
• Knowledge of large models and hyper-parameter optimization.
• Knowledge of high-performance computing and its applications.
• An excellent record of productive and creative research, as demonstrated by publications in top peer-reviewed journals.
• Motivated self-starter with the ability to work independently and to participate creatively in collaborative and frequently interacting teams of researchers.

 

Special Requirements:

Applicants cannot have received their Ph.D. more than five years prior to the date of application and must complete all degree requirements before starting their appointment. The appointment length will be up to 24 months with the potential for extension. Initial appointments and extensions are subject to performance and availability of funding.

 

For employment at Oak Ridge National Laboratory (ORNL), a Real ID compliant form of identification will be required. Additionally, ORNL is subject to Department of Energy (DOE) access restrictions. All employees must also be able to obtain and maintain a federal Personal Identity Verification (PIV) card as mandated by Homeland Security Presidential Directive 12 (HSPD-12) and Department of Energy (DOE) Order 473.1A, which requires a favorable post-employment background investigation.

 

To obtain this creden