Hurum Maksora
Tohfa

PhD student in Astrophysics · University of Washington

I am currently a graduate student at the University of Washington working at the intersection of data science and astronomy. Modern astronomy is drowning in data with simulations that take months of supercomputer time, telescopes generating terabytes every night, surveys with billions of sources. I build the machine learning and computational tools that make data tractable spanning from emulators that replace expensive simulations with millisecond predictions to neural networks that control telescope optics in real time.

What excites me most is understanding where information lives in data. I've shown that standard analysis methods leave orders of magnitude of cosmological information on the table. I've trained networks that discover which features matter for science without supervision. I care about representation, compression, and the question of what makes a measurement informative whether that measurement comes from a telescope or a neural network. Before UW, I studied physics and math at Bryn Mawr College and spent a summer at Microsoft Research teaching neural networks to identify mosquito species from wingbeat interference in optical sensors.

Curriculum Vitae htohfa@uw.edu
Hurum Maksora Tohfa
5Publications
174Citations
10+Courses TA'd
Research

What I'm working on

Both theoretical simulations and observational surveys now produce data that pushes the limits of traditional analysis. I develop machine learning tools across the astronomical pipeline — emulators that bypass expensive simulations, compression algorithms that preserve scientifically relevant structure, and deployed systems that deliver the optical quality required for precision science.

Deployed on-sky

Real-time Neural Optics for the Vera Rubin Observatory

with Prof. Andrew Connolly · UW · 2023–present

Rubin needs to correct its optics every 36 seconds. I built a CNN that predicts and corrects gravity- and temperature-induced aberrations from out-of-focus sensor images, achieving 30× faster inference than the baseline solver. The hardest part was a brutal sim-to-real gap — I solved it by treating it as a physics-informed learning problem rather than pure domain adaptation. Now deployed on-sky for LSST.

Active

Finding Faint Solar System Objects with Deep Learning

with Prof. Andrew Connolly · UW · 2026

KBMOD searches for faint moving objects by shifting and stacking images. The existing classifier failed on faint 5σ candidates — where the real discoveries live. I diagnosed a data preprocessing bug, redesigned the model with channel attention so it could learn which coadd methods matter at different signal-to-noise levels, and recovered thousands of novel trans-Neptunian object candidates.

Submitted

Emulating Cosmic Reionization

with Prof. Matthew McQuinn · UW · 2024–2026

Each high-resolution reionization simulation eats ~200,000 CPU-hours. I trained a neural emulator that takes simulation parameters as input and directly predicts the ionizing photon mean free path — bypassing N-body simulations entirely. The emulator runs in milliseconds with 1.6% error across four orders of magnitude, enabling parameter exploration that would otherwise be computationally intractable.

arXiv: 2602.03923 →
In development

Neural Compression for Survey Astronomy

with Prof. Andrew Connolly · UW · 2024–present

Rubin will generate 20 TB every night for ten years. I'm developing compression algorithms that learn to identify and preserve scientifically critical features without supervision — the network discovers what matters from data statistics alone, achieving 30× compression while maintaining 99% detection accuracy for sparse structures.

In development

PeriodicOrbit.jl

with Prof. Eric Agol & T. Chatchadanoraset · UW

Standard exoplanet methods can't uniquely determine both mass and eccentricity. I co-developed a Julia package that exploits periodic orbit theory to break this degeneracy — if a planetary system returns to its exact initial state, that constrains the physics. Led to a Co-I role on an accepted JWST Cycle 4 proposal to observe TRAPPIST-1.

Active

Baryon Streaming Velocity & Structure Formation

with Prof. Matthew McQuinn · UW · 2024–present

The relative velocity between dark matter and baryons after recombination is a subtle effect from the early universe that turns out to have big consequences for how and when the first cosmic structures form. Implementing the physics in MP-Gadget simulations to quantify the impact on halo mass distributions.

Full research details
Publications

Papers

All publications & presentations
Writing

Research stories, minus the jargon

Building Virtual Universes to Understand Reionization

How I use high-resolution simulations and neural networks to model what happened when the first stars lit up the universe, and why getting the small-scale physics right matters for everything else.

Read now →