Hi! I'm a Turing Scholar at UT Austin (Class of 2028).

I am interested in numerical methods for differential equations, and increasingly the machine learning side. I work across the Julia SciML ecosystem, with a focus on ODE solvers, and I'm building toward the intersection: scientific machine learning for dynamical systems and climate.

Currently at JuliaHub working with Chris Rackauckas, and researching machine learning methods for Hamiltonian dynamics with Chandrajit Bajaj at UT's Center for Computational Visualization.

• Adaptive-order Radau IIA stiff ODE solver (IEEE HPEC 2025, arXiv:2412.14362). Roughly 2x faster than the classic Fortran RADAU. Presented at IEEE HPEC and JuliaCon. Landed in OrdinaryDiffEq.jl: order adaptivity, parallelization.
• Discontinuity detection for ODE solvers (accepted, 2026 American Modelica Conference). A stepping controller that detects and handles discontinuities in ODE solvers; in collaboration with MERL, delivered up to 10× speedups on their vapor-compression cycle models. Shipped as an ecosystem-wide feature (#3121, #3720).

Some highlight projects:

• Butterfly Factorization in LinearSolve.jl #785 and RecursiveFactorization.jl #97.
• Authored SparseBandedMatrices.jl to support the above work.
• Symbolic logging spanning the stack: ModelingToolkit.jl #4720, SciMLBase.jl #1434, and OrdinaryDiffEq.jl #3731.
• Plus other contributions across the ecosystem.

Deep-Learning: the modern DL canon built from scratch (numpy backprop, then PyTorch), plus applied projects. This is my bridge from numerics into ML. I work through some of the traditional ML projects and put my own scientific twist on a few.

• Kolmogorov flow with neural ODEs: learn the dynamics of a chaotic 2D turbulent flow, first at a fixed viscosity, then conditioned on variable viscosities as an input.
• GNN interatomic potentials: learn a Lennard-Jones energy, get forces by autograd, run it as a differentiable MD sim.
• MRI to CT translation: six image-translation architectures benchmarked (U-Net through conditional diffusion).
• Acrobot control: from-scratch DQN and PPO, continuous swing-up and balance with an LQR handoff.

• Regeneron Science Talent Search Top 300 Scholar (2025), for the Radau research
• MIT PRIMES Computer Science Research Scholar (2024 to 2025)
• 2x National Science Bowl Champion (2023, 2024)
• Non-Trivial Fellow (~1% acceptance), on climate change and agricultural yields
• USESO National Finalist (2024, 2025)