Kurt Hamblin, Allison Kirkpatrick, Bren E. Backhaus, Gregory Troiani, Fabio Pacucci, Jonathan R. Trump, Alexander de la Vega, L. Y. Aaron Yung, Jeyhan S. Kartaltepe, Dale D. Kocevski, Anton M. Koekemoer, Erini Lambrides, Casey Papovich, Kaila Ronayne, Guang Yang, Pablo Arrabal Haro, Micaela B. Bagley, Mark Dickinson, Steven L. Finkelstein, Nor Pirzkal

We present AGNBoost, a machine learning framework utilizing XGBoostLSS to identify AGN and estimate redshifts from JWST NIRCam and MIRI photometry. AGNBoost constructs 121 input features from 7 NIRCam and 4 MIRI bands-including magnitudes, colors, and squared color terms-to simultaneously predict the fraction of mid-IR $3-30\,\mu$m emission attributable to an AGN power law (frac$_\text{AGN}$) and photometric redshift. Each model is trained on a sample of $10^6$ simulated galaxies from CIGALE providing ground truth values of both frac$_\text{AGN}$ and redshift. Models are tested against both mock CIGALE galaxies set aside for testing and 698 observations from the JWST MIRI EGS Galaxy and AGN (MEGA) survey. On mock galaxies, AGNBoost achieves $15\%$ outlier fractions of $0.19\%$ (frac$_\text{AGN}$) and $0.63\%$ (redshift), with a root mean square error ($\sigma_\text{RMSE}$) of $0.027$ for frac$_\text{AGN}$ and a normalized mean absolute deviation ($\sigma_\text{NMAD}$) of 0.011 for redshift. On MEGA galaxies with spectroscopic redshifts, AGNBoost achieves $\sigma_\text{NMAD}$ = 0.074 and $17.05\%$ outliers, with most outliers at $z_\text{spec} > 2$. AGNBoost frac$_\text{AGN}$ estimates broadly agree with CIGALE fitting ($\sigma_\text{RMSE} = 0.183$, $20.41\%$ outliers), and AGNBoost finds a similar number of AGNs as CIGALE SED fitting. The flexible framework of AGNBoost allows straightforward incorporation of additional photometric bands and derived qualities, and simple re-training for other variables of interest. AGNBoost's computational efficiency makes it well-suited for wide-sky surveys requiring rapid AGN identification and redshift estimation.