Sign in

gully

@gully.bsky.social
1.4K followers 78 following 233 posts

interpretable machine learning for atmospheric and astronomical data analysis, near-IR spectra, climate tech, stars & planets; bikes, Austin, diving off bridges into the ocean.

PostsRepliesMedia
gully @gully.bsky.social · 25/09/2026
oddly satisfying to see the defocused PSF
010
gully @gully.bsky.social · 03/07/2026
Well deserved, one of the coolest things I’ve seen in the last few years, without a qualifier
010
gully @gully.bsky.social · 19/06/2026
Must go to this!
011
gully @gully.bsky.social · 16/06/2026
Very cool paper! Maps out the starspots in exquisite detail, among the best constraints on spot physical properties anywhere. Exoplanets make great scanning reticles for stellar surface features, in addition to their other virtues.
030
gully @gully.bsky.social · 16/06/2026
Also curious how ML/AI fits in, the ML for Cool Stars splinter session in Toulouse feels so quaint now by comparison to modern standards, but at the same time methods like normalizing flows and differentiable models still feel underdeveloped, would be happy to hear developments in this area!
000
gully @gully.bsky.social · 16/06/2026
It was really cool seeing the EPRV and stellar communities come together at Cool Stars, in Uppsala, Boston, Toulouse. It felt like witnessing the evolution of two scientific communities learn with different techniques and motivations but same physics, curious how this year lands in that evolution!
110
gully @gully.bsky.social · 16/06/2026
FOMO on Cool Stars in Japan right now! What are the cool new talks? Best splinter sessions? Need to know!
110
gully @gully.bsky.social · 05/06/2026
Exoplanet atmospheres 🤝 starspots, nice, curious to see the tomographic analysis of the spots
010
gully @gully.bsky.social · 25/05/2026
CCA! 🤩
020
gully @gully.bsky.social · 05/05/2026
Would adore talking physics-based models in Sydney at NeurIPS! How to make the stars align…
020
gully @gully.bsky.social · 28/04/2026
I pitched Kepler/K2 advertising swag that simply said “THERE ARE SO MANY PLANETS” in all caps. There are!
020
gully @gully.bsky.social · 15/04/2026
Just did a double take looking at the size of that echelle grating, whoa
020
gully @gully.bsky.social · 27/03/2026
I think astronomers are particularly effective at the use (and acceptance) of pathological scenarios for the purpose of illustration and education. We don’t mind conjuring impossibly weird planets as what-if scenarios. I sense that other fields can view pathological examples as distracting.
000
Reposted by gully
Matthew Kenworthy @mattkenworthy.bsky.social · 26/03/2026
If you haven’t seen 3Blue1Brown videos yet, you are missing out! His latest one is on M.C. Escher’s picture frame distortion is absolutely breathtaking, using visuals to introduce complex analysis in what feels a very intuitive way. All his videos are excellent and well worth a browse.
M.C. Escher's The Picture Gallery shows a highly distorted image of a continuously zoomed image, and the conformal mapping to a plane with infinitely tiled copied of this distorted image.
084
gully @gully.bsky.social · 25/03/2026
A free, cool idea I was excited about at Kepler/K2: There exists a transformation to make nonstationary stellar variability into stationary. Deep Kernel Learning was the fad at the time, but I’m not sure anyone ever pulled this off successfully. The hope was to make a physics based transformation
020
gully @gully.bsky.social · 25/03/2026
Bayesian priors = “priced in”
010
gully @gully.bsky.social · 25/03/2026
Deeply heartwarming to see folks in the same offices, Alexa even made me a Cuban cafecito!
000
gully @gully.bsky.social · 25/03/2026
I just stopped by UT Austin to return some long lost inventory, and was delighted to see so many familiar faces and hear about the great work going on! 🔭 🪐 my grad school instrumentation memos are actively being used—surreal!
111
gully @gully.bsky.social · 10/03/2026
I was about to propose a feature then did YAGNI and stopped, where do I collect my software engineering award.
000
gully @gully.bsky.social · 10/03/2026
h/t CTVC Sightline newsletter
000
gully @gully.bsky.social · 10/03/2026
China 5 year plan includes “Aim to double non-fossil energy in 10 years”, already plateauing emissions ☀️💨⚡️🔋⚛️🌋
110
gully @gully.bsky.social · 26/02/2026
Talk from Earl Bellinger describing black holes in the interior of stars, and observational consequences: youtu.be/RW0GX36neKM?...
youtu.be
Earl Bellinger: Putting compact objects where they don’t belong...(October 24, 2025)
YouTube video by Flatiron Institute
000
gully @gully.bsky.social · 26/02/2026
I suppose this channel is rare enough that we’re not likely to see one on a nearby star in our lifetimes, but still it’d be cool. Also we’re not sure this channel exists.
100
gully @gully.bsky.social · 26/02/2026
The planets would still have IR, you could do direct imaging on them, even better if they were young, you’d probably want a high mass host star for the planets to be young enough. The BH removes the host star for you, no coronagraph needed
100
gully @gully.bsky.social · 26/02/2026
So there’s the prospect of planets that previously had a host star suddenly orbiting a black hole, but without the destructive phase of traditional HR diagram stellar evolution. Dark Planet systems.
100
gully @gully.bsky.social · 26/02/2026
Assumes you have some precursor to predict the accretion event timing, probably wouldn’t work out, free fall time is at least comparable to light echo time, etc. but upside bonus if they were eclipsing before the event, resulting BH could lens the planets if geometry is perfect?
100
gully @gully.bsky.social · 26/02/2026
New planet detection technique: Look for stars that disappear entirely (black holes in interior, cf Bellinger) then look for light echo plateaus from the reflected light of planets. Apparently ~100 M dwarfs have disappeared consistent with this mechanism. You’d have a few light minutes of signal.
100
gully @gully.bsky.social · 25/02/2026
Currently reading the Project Hail Mary book and there’s a ton of astrophysics in it—no spoilers but early on there’s infrared spectroscopy, ALMA, and more. Curious to hear what other astronomers think!
120
gully @gully.bsky.social · 25/02/2026
For astrometry, you’re right that better knowledge of the PSF yields benefits in the high SNR regime as well, since asymmetries in the PSF shape need to be understood to account for biases in the centroid. Chromatic effects mean the PSF fitting should help in all cases
100
gully @gully.bsky.social · 25/02/2026
Good point, I suppose it depends on the goal— for photometry, aperture-based methods give you comparable results to PSF fitting in the high SNR regime. In the low SNR regime, the background noise dominates the aperture based methods, so PSF fitting boosts delivered SNR by weighting the pixels.
200
gully @gully.bsky.social · 25/02/2026
Well said
010
gully @gully.bsky.social · 25/02/2026
Neat work on precision PSF modeling of Gaia! Little known fact: we captured Kepler driftscan PSFs during downtime in the K2 mission. PSF modeling is great boon for low SNR science, extended objects, etc.
110
gully @gully.bsky.social · 20/02/2026
Good band name, or maybe a softball team composed in part by punk rock astronomers with a fastball
010
gully @gully.bsky.social · 12/02/2026
It’s me
040
gully @gully.bsky.social · 10/02/2026
Sans serif math in LaTeX is so satisfying. \mathsf{}
000
gully @gully.bsky.social · 10/02/2026
Above all it’s just great to hear Dan’s voice, miss that guy
010
gully @gully.bsky.social · 10/02/2026
youtu.be/yUAZGR9RwPI?... @benjaminpope.bsky.social it’s mostly live coding with examples of JAX-isms and some new goodies like context managers for float64
youtu.be
Dan Foreman Mackey: JAX Training
YouTube video by Flatiron Institute
110
gully @gully.bsky.social · 10/02/2026
Very cool opportunity to make differentiable physics models for EPRV, work with Ben and co, and live in a cool place. Pretty much a dream job, apply!
010
gully @gully.bsky.social · 06/02/2026
Still want a JAX for science conference, where all my JAX scientists at
000
gully @gully.bsky.social · 06/02/2026
The sheer joy I felt just now discovering that there is a new Dan Foreman Mackey video about JAX.
3112
Reposted by gully
Paige Bailey (webpaige.dev) @dynamicwebpaige.bsky.social · 05/02/2026
standing room only for physics? faith in humanity restored ✨👏
0135
gully @gully.bsky.social · 04/02/2026
I still feel like differentiable models are under appreciated, it’s like a magic super power for interpretability and extensibility. Maybe LLM-assisted coding can bring autodiff to more science teams? Still a mental leap to adoption.
000
gully @gully.bsky.social · 04/02/2026
Everything becomes so much easier when you can forward model your data
000
gully @gully.bsky.social · 04/02/2026
Looking back on this talk about autodiffable spectral models, there is still so much promise, I’d love to see what folks have been up to in this space! speakerdeck.com/gully/blase-...
speakerdeck.com
blasé: An interpretable transfer learning approach to cool star échelle spectroscopy
Comparison of échelle spectra to synthetic models has become a computational statistics challenge, with over ten thousand individual spectral lines affe…
010
Reposted by gully
Brett Morris @brettmorr.is · 27/01/2026
New job opportunity at @stsci.edu: Senior Astronomical Data Scientist for data analysis tools, working closely with my team and me. This role is very similar to mine. If you have questions, I'm happy to answer them. 🧪🔭
recruiting2.ultipro.com
02417
gully @gully.bsky.social · 14/01/2026
Currently 83% carbon-free energy in Texas right now: - 40% wind - 33% solar - 10% nuclear Wholesale prices are negative, and 3.5 GW of batteries are charging. ☀️💨⚛️🌎
032
Reposted by gully
Jesse D. Jenkins @jessedjenkins.com · 04/01/2026
Despite headwinds, over 20 new EVs are coming to the US market in 2026, including several more affordable vehicles. Can any thrive without federal tax credits? insideevs.com/news/782572/... 🔌💡 🔌🚗
insideevs.com
Over 20 New EVs Are Coming In 2026. These Are The Seven I'm Most Excited About
We’ve been waiting for the next generation of EVs to arrive—cheaper, faster, better. Next year, they’ll finally be here.
4668
gully @gully.bsky.social · 16/12/2025
It’s wild that stellar limb effects start to matter at non-negligible fractions of a resolution element on the ELTs! Challenge… or opportunity!?
010
Reposted by gully
Jesse D. Jenkins @jessedjenkins.com · 25/11/2025
The first batteries rolled off the production line at Toyota’s new $14 billion battery plant in Liberty, North Carolina. The plant will produce batteries for hybrid, plug-in hybrid & electric models, which Toyota plans to make up 70% of its sales by 2030 electrek.co/2025/11/17/t... 🔌💡 🔌🚗
electrek.co
Toyota's 1,850-acre EV battery plant is live, but the company says this is just the start
Toyota’s $14 billion US battery plant is now up and running. The mega site spans 1,850 acres, or about 121...
0444
Reposted by gully
Chris Lintott @chrislintott.bsky.social · 12/11/2025
Get outside into a dark place if you can tonight - no guarantees, but a possible brilliant display of the northern lights.
920864