Applied mathematician and computational scientist
Postdoctoral Fellow, Engineering Sciences & Applied Mathematics, Northwestern University
I work at the intersection of scientific machine learning, dynamical systems, and numerical methods, with a focus on developing algorithms to discover real-world models. I design numerically stable, data-driven methods for discovering and solving differential equations and implement them as open-source software.
- 🔭 Currently: Postdoctoral Fellow at Northwestern University (with Dr. Niall Mangan), affiliated with the NSF–Simons National Institute for Theory and Mathematics in Biology (NITMB) and the Trienens Institute for Sustainability and Energy.
- 💼 Previously: Quantitative Analyst (AVP) at Citigroup, NYC (2021–2023), building large-scale C++ pricing and risk libraries for credit derivatives. PhD in Mathematics, University of Pittsburgh (2021).
- 📦 Maintainer of
dae-finder, a model-agnostic Python package for discovering differential-algebraic equations from noisy data. - 🤖 New for Fall 2026: teaching Agentic AI for Scientific Computing, a project-based graduate course at Northwestern on the principled, validated use of frontier AI agents (Claude Code, Codex/GPT, Gemini) in scientific computing.
- 🤝 Always open to new collaborators and interesting projects.
| Equation discovery from data | Stable, interpretable algorithms for learning differential-algebraic equations from noisy measurements: SODAs (Proc. Royal Society A, 2026), demonstrated on chemical reaction networks, the IEEE-39 power grid, and battery models. |
| Inverse problems & ill-conditioning | Why dictionary-based model discovery fails (diagnosed via inverse-problem theory) and how to fix it: QR-based library orthogonalization, multiple-shooting parameter estimation for stiff systems. |
| Multiphysics PDE solvers | Finite-element solvers for coupled Poisson–Nernst–Planck electrochemical systems; domain decomposition for Biot poroelasticity; MPI-parallel implementations. |
| Agentic AI for scientific computing | Protocols, validation frameworks, and reusable agentic skill sets so AI-assisted scientific computing is verifiable, reproducible, and accessible on lower-cost models. |
Poroelastic flow simulated with my MPI-parallel solver BiotDD
| Project | What it is | Stack |
|---|---|---|
| DAE-FINDER (PyPI) | Model-agnostic package for discovering differential-algebraic equations from data via sparse optimization; scikit-learn-compatible .fit()/.score() interface |
Python |
| FluidLearn | Physics-informed neural networks for fluid-flow PDEs, packaged for domain scientists | Python, TensorFlow/Keras |
| MMMFE-ST-DD | Parabolic-PDE solver using space-time multiscale mortar mixed finite elements with non-matching subdomain grids | C++, deal.II |
| BiotDD | Poroelastic flow simulator using MPI-based non-overlapping domain decomposition | C++, MPI, deal.II |
| deal.II | Contributor to the widely used open-source C++ finite element library | C++ |
- M. Jayadharan, N. M. Mangan, et al., "SODAs: Sparse Optimization for Discovery of Differential-Algebraic Systems from Data," Proc. Royal Society A, 2026. DOI
- M. Jayadharan, I. Yotov, "Multiscale mortar mixed finite element methods for the Biot system of poroelasticity," Comput. Methods Appl. Mech. Engrg., 2025.
- M. Jayadharan, M. Kern, M. Vohralík, I. Yotov, "A space-time multiscale mortar mixed finite element method for parabolic equations," SIAM J. Numer. Anal., 2023.
- Y. Feng, N. M. Mangan, M. Jayadharan† (senior author), "Ill-conditioning in dictionary-based dynamic-equation learning," under review at SIAM J. Life Sciences, 2026.
Full list on my website and Google Scholar.
Languages: Python · C++ · Julia (previously MATLAB, Fortran) Scientific ML: SINDy-family methods · neural ODEs · PINNs · sparse optimization Numerical methods: FEM · domain decomposition · space-time methods · multiple shooting FEM / HPC: deal.II · FEniCS · FreeFem++ · MPI · Slurm Data science: NumPy · Pandas · SciPy · SymPy · scikit-learn · TensorFlow/Keras