Marine mussels represent a valuable source for discovery of novel biologically active peptides targeting key enzymes in the coagulation system. However, traditional iterative isolation methods are often unsustainable for rare deep-sea mussels. To address this challenge, we introduced an integrated strategy that combines traditional enzymatic hydrolysis and bioassay-guided fractionation of Gigantidas platifrons protein hydrolysate (GPk) with AI-assisted virtual screening. By leveraging AlphaFold3 and stringent metrics, we achieved efficient identification while minimizing the use of scarce biological materials. This multi-stage approach yielded a novel FXIa inhibitor KDETRTPEEL. This peptide demonstrated moderate micromolar inhibition potency in vitro toward FXIa with an IC50 of 0.40 ± 0.02 mM, and exhibited favorable antithrombotic activity in vivo. Computational structural analysis indicated that the peptide forms stable hydrogen bonds and salt bridge interactions with key residues Arg38, Asp189, and Lys192 within the S1 and S2' pockets of FXIa, primarily driven by electrostatic interactions. This research highlighted a sustainable and practical paradigm that integrates AI-assisted virtual screening with conventional bioassay-guided fractionation, enabling accelerated exploration of deep-sea bioresources.