Group & Opportunities
A question you can actually finish
I advise SDSU Astronomy undergraduate and Master's students. Differentiable, modular software turns a big question into a focused one — so you can own a real project end to end rather than hold a corner of someone else's.
What you'd work on
The Jaxstro ecosystem is deliberately modular. Because each package owns one scientific stage and exposes clean, differentiable interfaces, a single well-scoped question — a new IMF prescription, a dynamical benchmark, a survey systematic, an identifiability test — becomes a self-contained project with a beginning and an end.
That is the point: many compact, finishable contributions rather than a few monolithic ones, on tools you can actually read.
Where to start looking: the explorables to see the models running, the software for what each package does, and the jaxstro and progenax documentation if you want to read the interfaces themselves.
Research interests
Recommended skills
None of this is a checklist you must clear before talking to me. It is what makes the first month go well.
- You can write and debug your own Python — not only run someone else's notebook.
- You are comfortable thinking in arrays: NumPy, or the willingness to learn it quickly.
- A terminal and
git, or no fear of picking them up. - Helpful, not required: JAX, matplotlib, LaTeX, any experience on a cluster.
You do not need prior research experience, and you do not need astrophysics beyond the introductory sequence. If you have written code for a class or a project and had to make it actually work, that is the relevant experience.
Before you email me
Have a look around first, then tell me three things. Two or three sentences each is plenty — I'm not looking for a polished proposal, I'm looking for evidence that something specific here caught your interest.
- Which package or question you'd like to work on, and what drew you to it.
- Where you are: your year, your major, and which of my courses you've taken, if any.
- One thing you'd want to try — a plot you'd make, a test you'd run, a number you'd check. It doesn't have to be right.
Please put SDSU-ASTRO in the subject line.
Potential collaborators
The software is open and built to be extended. If you work on star clusters, stellar populations, or survey inference and want to run new models, test a prescription against your own, or design an observing program around what these models predict, I'd like to hear from you — the three items above don't apply.