Master's Thesis — Neural Solvers for Mixed-Integer Nonlinear Programming

Master's Thesis Neural Solvers for Mixed-Integer Nonlinear Programming

Mixed-integer nonlinear programs (MINLPs) are computationally challenging and remain difficult to solve at speed and scale. In this master thesis, the candidate will investigate neural optimization approaches that learn to generate near-optimal solutions to MINLPs significantly faster than conventional optimization algorithms, while respecting hard physical and engineering constraints. In particular, the thesis will investigate end-to-end surrogate model workflows that combine deep neural networks to solve the combinatorial (MILP) and continuous (NLP) problem in sequence. The NLP part will build on our GENCO model, by the extension of feasibility restoring layers. The proposed concepts will be demonstrated and benchmarked on electric power systems, for topology optimization under AC optimal power flow.

The ideal candidate will bring a strong foundation in machine learning, optimization, or applied mathematics, and the motivation to advance the algorithmic frontier on industrially relevant large-scale optimization problems.

Why Join Us?

• You will join a world-class research environment and collaborate closely with researchers in AI, optimization, and energy systems.
• You will gain access to state-of-the-art machine learning frameworks, optimization software, and high-performance computing resources.
• Our environment encourages innovation and supports scientific publications and intellectual property creation.

Preferred Qualifications

• Bachelor’s degree in computer science, electrical engineering, mathematics, operations research, or a related field.
• Background in one or more of the following domains: machine learning/AI, optimization, mathematical programming, power systems, control, or applied mathematics.
• Strong programming skills in Python and PyTorch; experience with scientific computing and deep learning on compute clusters.
• Experience with optimization frameworks (e.g., Gurobi, Pyomo, JuMP, IPOPT) is a plus.
• Self-motivated, team-oriented, and eager to work on interdisciplinary research problems.

Diversity & Work Environment

IBM is committed to fostering diversity and inclusion in the workplace. You will join an open, multicultural research environment that values different perspectives and supports flexible working arrangements. Our goal is to help all genders and backgrounds thrive professionally while maintaining a healthy work–life balance.