Understanding molecular determinants for protein interface formation has been an ongoing challenge in the field of protein design. Here, we try to combine physics and machine-learning based approches to design protein binders with tailored properties and binding sites. We use RFdiffusion, LigandMPNN, BindCraft and iterative AF3 and Boltz-2 refolding circles to identify the most promising binder candidates for experimental validation. For scoring, we included the ipSAE, ipTM, interaction energies and a desolvation penalty.