Predicting biomolecular interactions is fundamental to rational drug design, yet achieving experimentalaccuracy and generalisability across novel chemical space remains a critical bottleneck. While deeplearning approaches such as AlphaFold 3 have advanced structure prediction, benchmarks reveal thatlimitations persist in generalising to unexplored regions of molecular space, estimating binding affinity,and detecting molecular binding sites on previously uncharacterised protein surfaces. Here, we introducethe Isomorphic Labs Drug Design Engine (IsoDDE), a unified computational system designed to addressthese limitations. We demonstrate that IsoDDE more than doubles the accuracy of AlphaFold 3 on achallenging protein-ligand generalisation benchmark, successfully modelling complex out-of-distributionevents such as induced fits, and accurately identifying novel binding pockets. In the biologics domain,IsoDDE substantially outperforms existing models, providing a new state of the art for antibody-antigeninterface prediction and CDR-H3 loop modeling. Finally, for small molecule binders, IsoDDE’s affinitypredictions exceed gold-standard physics-based methods, bridging the gap towards experimental-gradeprecision without the computational overhead of traditional physics-based workflows. Our resultsdemonstrate that IsoDDE offers a scalable foundation for AI drug design, providing the predictive fidelityrequired to navigate novel biological systems with unprecedented accuracy.