Data Scientist

You will build our in-house simulation stack - the models that predict what our fusion machines will do, and the inference machinery that pulls physics out of shot data. This is not a “fit a curve to a dashboard” data science role., * Build a coupled simulation framework for our DPF: pulsed-power circuit, sheath formation and run-down (snowplow / slug / Lee-type models, progressing toward MHD), pinch and instability development, and neutron and X-ray production.

• Implement radiation source-term models: thermonuclear vs. beam-target neutron-yield decomposition, bremsstrahlung and line-emission spectra, and anisotropy.
• Develop surrogate and reduced-order models so designers and physicists can iterate on parameter sweeps without standing up an HPC job each time.
• Build the data pipeline that ingests every shot’s diagnostic stream, including Rogowski coils, B-dots, silver activation, time-of-flight neutron detectors, filtered diodes, and fast cameras, and joins it to predicted output for systematic residual analysis.
• Quantify uncertainty seriously: Bayesian inference over model parameters, identifiability analysis, and honest error bars on yield predictions.
• Stand up simulations of supporting machine outputs, including anode/cathode lifetimes, electrode erosion models, and capacitor-bank aging, so operations decisions are informed by physics rather than vibes.
• Publish internally with the same standards you’d publish externally: derivations written out, assumptions stated, code reviewed.
• Eventually grow a small simulation team. For now, you will be a force multiplier of one., * Take responsibility. If something is broken and you can fix it, fix it. See things through.
• Move fast. Make decisions, learn quickly, and keep moving forward.
• Think bigger. Simplify relentlessly. Don’t default to the safe answer.
• Disagree and commit. The best idea wins. Once a decision is made, we move.
• Desire to win. This is hard. Expect intensity. We are here to win.
• Plan B is to make Plan A work.

• PhD in Physics, Applied Mathematics, Plasma Physics, Computational Science, Nuclear Engineering, or a closely related field. An exceptional Master’s candidate with a strong publication record will be considered.
• Strong academic record, with a target GPA of 3.8+ from a competitive program.
• Demonstrable depth in at least one of: magnetohydrodynamics, kinetic plasma theory, radiation transport, or pulsed-power circuit modeling. Familiarity with the others.
• Fluency in vector and tensor calculus, Maxwell’s equations in arbitrary geometries, hyperbolic PDEs and numerical schemes, Bayesian statistics, and optimization under constraints.
• Production-grade Python using NumPy, SciPy, xarray, JAX, or PyTorch for differentiable physics, with comfort dropping into C++ or Fortran when Python isn’t fast enough.
• Experience writing simulation code from scratch, not just running someone else’s solver.
• Ability to read a physics paper, identify the load-bearing assumption, and determine quickly whether it applies to our regime., * Hands-on experience with one or more of: MCNP, GEANT4, FLUKA, PIC codes such as LSP, EPOCH, or WarpX; MHD codes such as USim, FLASH, or HYDRA-class; or DPF-specific codes such as Lee model or GORGON-adjacent work.
• Prior work fitting models to noisy experimental diagnostics, including neutron TOF, X-ray spectroscopy, or magnetic-probe arrays.
• HPC experience including SLURM, MPI, GPU acceleration, and profiling at scale.
• Familiarity with the DPF literature or adjacent fields such as Z-pinch, MagLIF, or ICF, and a track record of getting up to speed quickly.
• Strong software-engineering hygiene: version control, tests for numerical code, and reproducible environments.

• Medical, dental, and vision coverage.
• Relocation assistance for roles that require it.
• Flexible time off. Take what you need, no accrual tracking.
• 2 weeks of paid time off built into the end of each year, subject to team and business needs.
• Supportive leave of absence policies.
• Paid leave for new parents.