Spectroscopic surveys like the Dark Energy Spectroscopic Survey (DESI) and Euclid are mapping the spatial distribution of millions of galaxies, with Emission Line Galaxies being the dominant tracer between redshifts $0.8<z<1.6$. Standard approaches for extracting cosmological information from the clustering of these galaxies typically discard highly
constraining measurements from the nonlinear regime. In this talk, I will present the first cosmological constraints derived from the clustering of ELGs from the DESI OnePercent data release using scales between 0,3 Mpc/$h$ and 40 Mpc/$h$. The constraints are obtained using the SHAMe-SF model, a modification of subhalo abundance matching
specifically designed for star-forming samples. The inference pipeline was validated using mock samples from the MillenniumTNG simulation, yielding unbiased results for a DESI-like sample and an H$\alpha$-like sample. We can constrain $\sigma_8$ to 10% and $\Omega_{\rm{m}}h^2$ to 12%, yielding values consistent with Planck and the full-shape
analysis of DESI combining all tracers.