We use the dynamical algebra of a quantum system and its dynamical invariants to inverse engineer feasible Hamiltonians for implementing shortcuts to adiabaticity. These are speeded up processes that end up with the same populations than slow, adiabatic ones. As application examples we design families of shortcut Hamiltonians that drive two and a three-level systems between initial and final configurations imposing physically motivated constraints on the terms (generators) allowed in the Hamiltonian.