I am a machine learning research engineer. My training is in mechanical engineering and fluid physics, and I have spent the last several years on the seam between that world and this one — building models that learn from data without quietly abandoning the physics they are supposed to obey.

The short version: neural networks are very good at fitting things and very bad at knowing when they are wrong. Physical law is a way of telling them. Most of my research has been some variation on that theme, applied to complex fluids — the materials that are neither quite solid nor quite liquid and refuse to behave.

I live in Dallas. I write here when something will not leave me alone until I have put it in sentences.

Record

• 2026 — now Senior Research Engineer, Machine Learning — APQX Agentic and modeling systems for industrial R&D. Dallas.
• 2024 — 2026 Research Engineer, Machine Learning — APQX Physical discovery models, from the lab to production.
• 2021 — 2024 PhD, Northeastern University Physics-guided machine learning for complex fluids. Boston.
• 2017 — 2020 MSc, K. N. Toosi University of Technology Coupled magneto-hydrodynamics; ferrofluid micro-mixing. Tehran.
• 2012 — 2017 BSc, K. N. Toosi University of Technology Mechanical engineering. Tehran.

Selected work

• Ongoing Agentic pipelines for physical systems Industrial-scale agents that act on real processes rather than describe them.
• 2024 Neural operators for rheological models Learning whole families of constitutive models. Journal of Rheology.
• 2024 Multi-fidelity neural networks Characterizing a fluid from very little data. Journal of Rheology.
• 2022 — RhINNs Rheology-informed neural networks. Open source.

Elsewhere

• Email saadat.m@northeastern.edu
• GitHub milowsa
• Scholar Publications
• LinkedIn m-saadat