Atom interferometers rely on prescribed sequences (splitting, reflection, recombination), optimized step by step, whose precision improves by repeating measurements. This project asks a complementary question, crucial when interrogation time or the number of runs is limited: how much information can a single experimental cycle deliver? With a rubidium Bose–Einstein condensate in an optical lattice whose phase is modulated in real time, we recently showed that a modulation optimized from the Fisher information spontaneously builds an interferometer that reaches the quantum bound for force sensing in a fraction of a millisecond. The goal is now to identify what replaces the standard interferometric scheme when several parameters are measured simultaneously, such as the components of an acceleration, or an acceleration and a rotation. The quantity to optimize becomes a Fisher matrix, which encodes both the attainable precisions and the correlations between estimates. The project will first address force gradients, then two-dimensional vectorial sensing, comparing square and genuinely non-separable triangular lattices, and finally robustness against poorly known parameters and the role of atomic interactions. Combining experiment and numerical optimal control, it offers a complete training at the interface between cold atoms, quantum control and quantum metrology.