This project proposes PeptiCraft, an AI-driven platform for modular multifunctional peptide design. The system integrates multi-task prediction models, linker-aware fusion generation, and delivery-constrained optimization to address limitations of current single-function peptide design approaches. By combining machine learning with rule-based feasibility filtering, the platform prioritizes fusion peptides with improved functional compatibility, physicochemical plausibility, and application suitability. Unlike conventional pipelines, PeptiCraft jointly considers bioactivity, linker geometry, and intracellular delivery within a unified framework. The project aims to establish a practical candidate discovery workflow that supports downstream experimental validation under data-limited conditions.
