Presently, noisy intermediate-scale quantum computers encounter significant technological challenges that make it impossible to generate large amounts of entanglement. We leverage this technological constraint as a resource and demonstrate that a shallow variational eigensolver can be trained to successfully target quantum many-body scar states. Scars are area-law high-energy eigenstates of quantum many-body Hamiltonians, which are sporadic and immersed in a sea of volume-law eigenstates. We show that the algorithm is robust and can be used as a versatile diagnostic tool to uncover quantum many-body scars in arbitrary physical systems.