This study proposes a multi-criteria virtual screening process for the rapid discovery of selective estrogen receptor degraders (SERDs) to degrade estrogen receptor alpha (ERα). Highly represented (70%) in breast tumors, ERα is responsible for origin and progression of the cancer, serving as a principal molecular target for therapeutic intervention. The multi-tiered screening method incorporates Glide docking, a physics-based prediction program, Karmadock, a deep-learning-based ligand docking model, and Carsidock, a deep-learning based computational algorithm for binding-conformation prediction. The biological activity of the selected compounds were tested by in-vitro cell proliferation assays using ERα-positive breast cancer cell lines. A library of 330,000 molecules was first filtered via Lipinski's rule of five, with the 67,916 compounds remaining were delivered to the multi-tiered screening framework. The final stages of screening yielded 21 candidates predicted to have favorable pharmacokinetic properties and binding affinities. The proliferation assays of four of the SERD candidates demonstrated potent inhibitory effects and strong binding affinities, supporting their potential as ERα inhibitors and validating the developed screening method as efficient and reliable. Overall, the screening framework presented in the paper provides a broadly applicable, efficient strategy for discovering novel SERDs and holds promise for advancing therapeutic development against ERα-positive breast cancer.