We　propose　a　scheme　to　realize　error-resistant　nonadiabatic　binomial-code　geometric　quantum　computation　using　reverse　engineering.　A　strong　Kerr　nonlinearity　restricts　the　evolution　in　a　computational　subspace　of　the　binomial　code　and　a　two-photon　squeezing　drive　provides　the　connections　between　the　logical　states.　The　effective　Hamiltonian　possesses　SU(2)　dynamic　structure　and　is　analyzed　through　reverse　engineering　based　on　a　dynamic　invariant.　By　combining　reverse　engineering　with　the　optimal　control　method,　we　find　the　evolution　paths　for　nonadiabatic　geometric　quantum　computation　and　derive　the　control　field　robust　against　the　systematic　error.　Numerical　simulations　show　that　the　scheme　holds　excellent　resistance　to　the　systematic　error　and　is　still　well　implemented　in　the　presence　of　resonator　leakage　with　the　current　superconducting　nonlinear　resonator　technology.　Therefore,　the　scheme　may　provide　a　promising　approach　for　accurate　nonadiabatic　binomial-code　geometric　quantum　computation.　©　2022　Optica　Publishing　Group.