Member of Technical Staff, Scientific Data Engineering London or San Francisco

About Zenithon

Zenithon is an applied frontier lab for extreme physics.

We build foundation models that simulate plasmas, so engineers can explore far more of the design space before they build anything.

The role

Getting experimental and simulation data into a form a model can learn from is hard, slow work, and this role owns it.

You’ll take data from simulations, plasma diagnostics and the datasets our partners bring, clean it, pull out what matters and build datasets our ML researchers can rely on.

You’ll decide how we do this across instruments and domains, and you’ll automate as much of it as you can.

• What you’ll do
• Take in raw data from simulations, from instruments such as optical emission spectrometers, laser diagnostics, Langmuir probes and mass spectrometers, and from partner datasets in whatever form they arrive.
• Clean and validate it: calibrations, baselines, artifacts, missing values, and streams recorded on different clocks.
• Extract the quantities that matter, with their uncertainties, working closely with our physicists and ML researchers.
• Build training-ready datasets in Zarr and similar formats, with clear schemas, metadata and versions.
• Automate the path from raw files to datasets so new data flows through without hand-holding, at terabyte scale and beyond.
• Help decide what gets recorded in experiments, and work with our ML engineers on how data feeds training.
• What you’ll bring
• A degree in physics, chemistry, astrophysics, engineering or another science. Many people in this role have a Ph D, but it isn’t required.
• Hands-on experience taking messy experimental or simulation data to clean, analyzed results, ideally data you or your lab produced.
• Strong Python and the scientific stack (Num Py, Sci Py, pandas, xarray), with Git and automated tests.
• Experience with large datasets and the tools to handle them.
• Clear, fluent English, written and spoken, for working closely with physicists, engineers and ML researchers.

You don’t need years of industry experience. We care most about what you’ve built and how fast you learn.

• Bonus points
• You’ve trained ML models on data you processed yourself, for example to classify images, segment scans or identify compounds in spectra.
• Terabyte- to petabyte-scale data from experiments such as the LHC, synchrotrons or sky surveys.
• Spectroscopy, mass spectrometry, laser diagnostics such as Thomson scattering, or imaging data.
• Zarr, HDF5, Net CDF or ROOT; Dask, Ray or Spark; workflow tools and cloud storage.
• Signal processing and fitting: baselines, peak fitting, deconvolution and uncertainty propagation.
• You’ve built tools others rely on, contributed to open source, or worked at a startup.
• How we work

We’re a small team that works hard, moves fast and holds a high bar.

Two principles guide everything we do: intellectual ambition in what we attempt, and empirical precision in how we measure it.

You’ll own real problems from your first week, and your work will shape what the company becomes.

What we offer

Competitive salary and meaningful equity.

You can be based in London, at our office near St Paul’s, or in San Francisco, on site at least three days a week.

You must be authorized to work in the UK or the US, depending on where you’re based.

Some projects may involve export-controlled data, which can require additional screening.

Zenithon is an equal opportunity employer. We welcome applicants of every background.