Archis Joglekar
Lead Simulation Intelligence Engineer - Nuclear Fusion
Seattle, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolArchis Joglekar is a Lead Simulation Intelligence Engineer with a decade of experience building high-performance, differentiable simulation platforms for inertial fusion and plasma physics. He combines deep expertise in computational physics, neural differential equations (contributing bug fixes and test tooling to the torchdyn library), and distributed/HPC engineering to deliver end-to-end workflows that include gradients, data-driven models, and optimization loops. As founder and principal investigator at Ergodic LLC and collaborator with top labs and universities, he designs scientific data infrastructure and cloud-scale pipelines for data-intensive experiments. Equally comfortable in CUDA, MPI, PyTorch and JAX, he bridges research and production to accelerate physics discovery and experiment interpretation. A not-obvious strength: he repeatedly translates complex legacy simulation codes into modern, differentiable, and testable systems that double performance while enabling new physics via gradients.
10 years of coding experience
12 years of employment as a software developer
Doctor of Philosophy (Ph.D.) Nuclear Engineering, Doctor of Philosophy (Ph.D.) Nuclear Engineering at University of Michigan
English, Marathi, German, Hindi
Github Skills (8)
debugging10
pytorch10
debug10
deep-learning10
ode10
numerical-methods10
neural10
python10
Programming languages (9)
Github contributions (5)
DiffEqML/torchdyn
Feb 2022 - Sep 2022
A PyTorch library entirely dedicated to neural differential equations, implicit models and related numerical methods
Role in this project:
ML Engineer
ML EngineerContributions:2 reviews, 46 commits, 10 PRs in 7 months
Contributions summary:Archis primarily focused on debugging and improving the `torchdyn` library, which is centered around neural differential equations and related numerical methods. Their contributions involved fixing bugs in the higher-order forward layer and notebooks, particularly those related to the behavior of the NODE and the need to restructure the training step. They also implemented a dummy solver for testing and integrated the `save_at` functionality in the `odeint` function. These changes improve the core functionality and testing capabilities within the library.
dynamical-systemsdifferentialdedicateddifferential-equationsnumerical-methods
ergodicio/adept
Jun 2023 - Jan 2025
Automatic-Differentiation-Enabled Plasma Transport in JAX
Contributions:5 releases, 4 reviews, 143 PRs in 1 year 7 months
automatic-differentiationboltzmann-equationcomputational-fluid-dynamicsfluid-mechanicskinetic-modeling
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.