The cosmic dawn and dark ages represent the largest unexplored epochs in the history of the Universe. Observations of the redshifted 21-cm signal from neutral hydrogen offer a unique opportunity to probe the formation of the first stars and galaxies while simultaneously testing the fundamental nature of dark matter. Several ongoing and upcoming experiments, including SARAS, PRATUSH, REACH, and the SKA, are expected to deliver increasingly sensitive measurements over the coming decade.

Realising the full scientific potential of these observations requires more than instrumental sensitivity. The interpretation of 21-cm measurements is currently limited by theoretical uncertainties in radiative transfer modelling, challenges in extracting a faint cosmological signal from bright foregrounds, and the need for statistically robust inference frameworks capable of combining multiple observational probes.

In this talk, I will present a programme aimed at developing bias-controlled modelling and inference tools for 21-cm cosmology. I will discuss recent advances in Lyman-α radiative transfer simulations and their impact on predicted cosmic dawn signals, introduce the ECHO21 framework for modelling global 21-cm observations from the dark ages to reionization, and describe Bayesian approaches for constraining both astrophysical processes and exotic dark matter scenarios. I will also highlight ongoing efforts toward robust signal extraction for experiments such as SARAS and REACH, and discuss opportunities for joint analyses of 21-cm observations and high-redshift galaxy measurements from JWST. Together, these developments provide a pathway toward transforming forthcoming 21-cm measurements into reliable constraints on the first stars, early galaxies, and the nature of dark matter.