Master Thesis: Neural Optimization for Mixed-Integer Nonlinear Programs (MINLPs)

IBM Research Zurich seeks a Master's student to develop neural optimization methods for mixed-integer nonlinear programs, focusing on electric power system applications.

This internship at IBM Research Zurich centers on exploring neural optimization techniques to solve mixed-integer nonlinear programs (MINLPs) more efficiently than traditional algorithms. The project aims to create end-to-end surrogate model workflows that integrate deep neural networks to address combinatorial and continuous challenges. Success will be validated through benchmarking on electric power systems, particularly regarding topology optimization under AC optimal power flow constraints.

Responsibilities

• Explore neural optimization strategies tailored for MINLPs.
• Develop end-to-end surrogate model workflows that merge deep neural networks.
• Enhance the GENCO model by adding feasibility-restoring layers for the nonlinear programming component.
• Test and benchmark the proposed concepts using electric power system scenarios.

Requirements

• Solid background in machine learning, optimization, or applied mathematics.
• Drive to push the boundaries of algorithmic solutions for large-scale industrial optimization challenges.
• Proficient programming abilities in Python and PyTorch.
• Hands-on experience with scientific computing and deep learning tasks on high-performance compute clusters.
• Self-driven attitude with strong teamwork skills and interest in interdisciplinary research.

Nice to have

• Prior exposure to machine learning, AI, optimization, mathematical programming, power systems, control theory, or applied mathematics.
• Familiarity with optimization frameworks such as Gurobi, Pyomo, JuMP, or IPOPT.

What the company offers

• Opportunity to work within a world-class research setting alongside experts in AI, optimization, and energy systems.
• Access to cutting-edge machine learning tools, optimization software, and high-performance computing infrastructure.
• A culture that fosters innovation and supports the creation of scientific publications and intellectual property.

About the company

IBM Research Zurich operates as a premier research institution dedicated to fostering innovation. The environment is designed to support scientific discovery and the development of new intellectual property.

• The institution is recognized as a world-class research environment.
• The workplace culture actively encourages innovative thinking and scientific advancement.