Selecting the optimal QEC code requires expertise in both hardware noise and multiple code families. We present an automated engine that takes raw error rates, classifies noise, detects bias (Z/X asymmetry ratio) and correlated burst errors, and recommends the optimal code with minimum distance and resource estimate. The decision tree maps Z-biased noise to repetition codes, correlated burst noise to color codes, very low rates to Steane codes, and general noise to surface codes. Integration with FTQC resource estimation provides total physical overhead. We demonstrate up to 52.5x reduction in physical qubit count compared to default surface code selection when noise exhibits strong bias.