Yeonsu Jung

Simulation · numerical optimization · real-to-sim-to-real

Yeonsu Jung

Research Associate · Applied Mathematics, Harvard SEAS · Mahadevan group

I simulate slender things in contact (whips, snakes, rod packings, filament grippers, printed lattices) and use optimization to make them do something on purpose, or to explain why they hold together.

Open to industry research roles in robotics and in applied numerical optimization. What I most want to work on is real-to-sim-to-real transfer: a simulator built from measurements of the real system, and results that still hold when they go back to it.

In the portfolio these run live: orbit them, scrub through time, compare runs.

Research interests

What I work on

  • Real-to-sim-to-real transfer

    Simulators built from measurements of the real thing and calibrated against hardware, so that what is optimized in simulation holds when it goes back.

    From an X-ray scan to a model

  • Physics simulation for robots

    Deformable and contact-rich bodies in the simulators robots train in: an exact elastic rod law for MuJoCo, and contact engines of my own in C++ and CUDA.

    Rods in joint coordinates

  • Numerical optimization

    Gradients through the simulation, Newton and Gauss–Newton solvers for equilibrium and inverse design, and energy minimization of thousands of bodies on a GPU.

    What is optimized, work by work

  • Control and learning

    Feedback that holds an unstable orbit, receding-horizon control, and reinforcement learning in thousands of parallel environments.

    A chain that learns to rear

  • Entanglement and grasping

    Why a pile of rods or a bundle of filaments holds together, and how to grasp with it: measured, explained, then optimized.

    Entangled rods and filaments

Basic information

At a glance

Now
Research Associate in Applied Mathematics, Harvard SEAS, with L. Mahadevan, since 2026.
Before
Postdoctoral Fellow at Harvard SEAS (2021–26), the Rowland Institute at Harvard (2020–21) and Seoul National University (2019–20).
Education
PhD, Mechanical Engineering, Seoul National University, 2019. MS, Biomedical Engineering, GIST, 2014. BS, Mechanical Engineering, POSTECH, 2012.
Tools
Python and JAX, C++17, CUDA. MuJoCo: C++ plugins, MJX, MuJoCo Warp.
Based in
Cambridge, Massachusetts.
CV
Download as PDF or read it here.

Selected papers

Four to start with

All publications