Microsoft Research, the research arm of Microsoft focused on long-term scientific and technological advances, has released a new AI-based system called OptiMind. It aims to address one of the most persistent barriers in operations research: converting real-world problems into mathematical models that can be optimized by machines.
Doug Burger, managing director of Microsoft Research Core Labs, confirmed the release on LinkedIn, positioning OptiMind as a way to translate plain language descriptions into formal optimization formulations such as mixed-integer linear programs. These models can be solved using established optimization engines.
Burger wrote that the system was built to address “the difficulty of formulating complex problems and systems in a way that can be optimized,” adding that OptiMind “translates natural language into mathematical formulations and facilitates the search for solutions with powerful optimization solvers.”
From natural language to supply chains and schedulers
According to Burger, OptiMind is designed for scenarios where systems are too complex, dynamic, or interconnected for manual modeling to scale. He said the system allows organizations to “optimize and improve complex systems such as supply chains, manufacturing systems, and global schedulers,” while supporting the testing and re-optimization of scenarios as conditions change.
This release reflects Microsoft Research’s broader push to combine large-scale language models with traditional optimization tools, rather than treating generative AI as a standalone solution. Berger said this approach allows users to continually explore alternatives as constraints, inputs, and goals evolve.
OptiMind is currently available for experimentation via Microsoft Foundry and Hugging Face, with benchmarks and data processing pipelines publicly available. Berger said the decision to make these assets public is aimed at supporting transparency and community-driven progress.
He said the system is part of a broader effort by Microsoft Research’s machine learning and optimization team to “democratize operational optimization with generative AI and agent solutions” that combine large-scale language models with optimization algorithms already used in simulators and industry.
Long-term ambitions beyond running a company
Burger also pointed to long-term applications that extend beyond enterprise workflows. He said the same technology could eventually be applied to larger systems such as cities, infrastructure and regional economies.
In his post, Berger wrote that tools like OptiMind could play a role in sustainability efforts, saying they would be “critical in reducing emissions and building a more sustainable future.”
He concluded by congratulating the Microsoft Research team behind the release and highlighting the contributions of researchers across optimization, machine learning, and system design.
