Machine learning decodes the universe's oldest light: what hydrogen absorption tells us

The Lyman-α forest is a thicket of absorption lines in quasar spectra that encodes information about dark matter, intergalactic gas, and the Universe's expansion history. Machine learning has transformed analysis by automating system detection, reconstructing quasar light, speeding up simulations by orders of magnitude, and enabling field-level inference that treats the forest holistically rather than absorption line by line. As new surveys produce torrents of data, these ML pipelines are becoming essential—not optional—for squeezing cosmological constraints from the signal.