Our de novo peptide design strategy for targeting RBX1 integrates sequence generation, novelty screening, structural evaluation, and energetic ranking into a unified pipeline to efficiently identify high-confidence binders. We first generate a diverse set of peptide candidates of varying lengths (5–40 amino acids) using PepMind, a multimodal discrete diffusion model that co-generates peptide sequences and 3D coordinates conditioned on the RBX1 target. To ensure novelty and compliance with competition rules, peptides are filtered against UniRef50 using MMseqs2 to remove sequences exceeding 25% identity to known proteins, with clustering analysis confirming high sequence diversity among retained candidates. The filtered peptides are then docked to RBX1 using RAPiDock to produce complex conformations, which are evaluated for energetic favorability using both Rosetta and FoldX, discarding weak binders. Top-ranking peptides are further assessed for structural reliability and interface quality with AlphaFold3 metrics (pLDDT, PAE, ipTM), and final selection prioritizes peptides that simultaneously optimize sequence novelty, binding energetics, and structural confidence. This integrated, multi-stage approach ensures that our design process balances chemical and structural optimization with strict novelty constraints, providing a rational and efficient path from initial generation to high-quality candidate selection.