Yahya Saleh

Hey, I’m Yahya, an applied mathematician and research engineer. I have experience in building and analyzing machine learning models for scientific computing and engineering problems.

I am genuinely passionate about pushing the boundaries of what is machine computable. I developed a new spectral-learning framework for solving Schrödinger equations that describe vibrations of molecules and built software that eases the adoption of this framework in practical applications (not that the framework is widely adopted but getting there :v).

In more industrial applications, I’ve built test automation and data-fixture tooling for banking systems, prototyped AI-agent integrations with the Model Context Protocol (MCP), and worked on cloud deployment of machine learning models.

Research interests: Approximation theory · Scientific machine learning · Statistical learning theory · Computational quantum molecular physics

Software

Active-Learning-of-PES — Active learning of potential-energy surfaces with regression-tree ensembles
FlowBasis — Spectral learning — basis sets augmented by normalizing flows for solving differential equations
vibrojet — Molecular rovibrational kinetic and potential-energy operators via Taylor-mode automatic differentiation

selected publications

Inducing Riesz Bases in L^2 via Composition Operators

Yahya Saleh and Armin Iske

Complex Analysis and Operator Theory, 2026

DOI arXiv Bib

@article{saleh2026riesz,
  title = {Inducing Riesz Bases in {$L^2$} via Composition Operators},
  author = {Saleh, Yahya and Iske, Armin},
  journal = {Complex Analysis and Operator Theory},
  volume = {20},
  pages = {21},
  year = {2026},
  doi = {10.1007/s11785-025-01874-5},
}

Computing Excited States of Molecules Using Normalizing Flows

Yahya Saleh, Álvaro Fernández Corral, Emil Vogt, Armin Iske, Jochen Küpper, and Andrey Yachmenev

Journal of Chemical Theory and Computation, 2025

DOI arXiv Bib

@article{saleh2025excited,
  title = {Computing Excited States of Molecules Using Normalizing Flows},
  author = {Saleh, Yahya and Corral, {\'A}lvaro Fern{\'a}ndez and Vogt, Emil and Iske, Armin and K{\"u}pper, Jochen and Yachmenev, Andrey},
  journal = {Journal of Chemical Theory and Computation},
  volume = {21},
  number = {10},
  pages = {5221--5229},
  year = {2025},
  doi = {10.1021/acs.jctc.5c00590},
}

Bounds on the Generalization Error in Active Learning

V. Menden, Yahya Saleh, and Armin Iske

In Proceedings of the 6th Northern Lights Deep Learning Conference (NLDL), 2025

arXiv Bib HTML

@inproceedings{menden2025bounds,
  title = {Bounds on the Generalization Error in Active Learning},
  author = {Menden, V. and Saleh, Yahya and Iske, Armin},
  booktitle = {Proceedings of the 6th Northern Lights Deep Learning Conference (NLDL)},
  series = {Proceedings of Machine Learning Research},
  volume = {265},
  pages = {168--175},
  year = {2025},
}

Active learning of potential-energy surfaces of weakly bound complexes with regression-tree ensembles

Yahya Saleh, V. Sanjay, Armin Iske, Andrey Yachmenev, and Jochen Küpper

The Journal of Chemical Physics, 2021

DOI Bib Code

@article{saleh2021active,
  title = {Active learning of potential-energy surfaces of weakly bound complexes with regression-tree ensembles},
  author = {Saleh, Yahya and Sanjay, V. and Iske, Armin and Yachmenev, Andrey and K{\"u}pper, Jochen},
  journal = {The Journal of Chemical Physics},
  volume = {155},
  number = {14},
  pages = {144109},
  year = {2021},
  doi = {10.1063/5.0057051},
}

Convergence theory for Hermite approximations under adaptive coordinate transformations

Yahya Saleh

2026

arXiv Bib

@misc{saleh2026hermite,
  title = {Convergence theory for Hermite approximations under adaptive coordinate transformations},
  author = {Saleh, Yahya},
  year = {2026},
  archiveprefix = {arXiv},
  primaryclass = {math.NA},
}