International Association
for Cryptologic Research

We are looking for a 2-years postdoc on Neuro-Symbolic learning in secure multi-party computation. The Villum Experiment research project “Neuro-Symbolic Federated Learning with Secure Multi-Party Computation” aims to explore the feasibility of training neural networks with logical constraints using secure MPC. The project addresses critical domains such as finance and healthcare, where data privacy is paramount and traditional data sharing is not an option. The research will focus on: Investigating differentiable logics (DLs) such as DL2, fuzzy logics, and logics of the Lawvere quantale, to evaluate their tractability and numerical stability under MPC frameworks, with formal correctness guarantees. Developing novel multi-valued logics tailored for MPC if existing ones prove inadequate. Implementing and benchmarking neuro-symbolic models trained under secure MPC protocols. The postdoc will: Conduct theoretical and empirical research on DLs and MPC. Develop prototype implementations using existing MPC frameworks or custom solutions. Collaborate across disciplines including cryptography, machine learning, logic, formal methods. Contribute to publications in top-tier venues and help shape a new research frontier. We seek a candidate with: A PhD in Computer Science, Mathematics, Data Science, or a related field. Strong background in at least some of the following: machine learning, logic, cryptography and secure multi-party computation, formal verification. Experience with federated learning, differentiable programming, or symbolic AI is a plus. Proficiency with various programming languages such as Python, C++/Rust, functional languages. Experience with interactive theorem provers such as Rocq, Lean or Isabelle is a plus. Ability to work independently and collaboratively in an interdisciplinary environment.

Closing date for applications:

Contact: Alessandro Bruni

More information: https://candidate.hr-manager.net/ApplicationInit.aspx?cid=119&ProjectId=181828&DepartmentId=3439&M