For the analysis of a mixture of organic molecules in solution, Nuclear Magnetic Resonance (NMR) is, with mass spectrometry, one of the two most used analytical techniques. NMR is often considered to be more quantitative, more reproducible, and more able to identify a solute. However, it lacks sensitivity and resolution. The sensitivity can be increased by employing the particular properties of parahydrogen to create a so-called hyperpolarized state, which transiently but considerably increases the NMR signal. Regarding resolution, it can be notably improved by the use of multidimensional NMR spectroscopy, which should be fast in the case of the analysis of hyperpolarized species. Liquid-liquid extraction, a very frequently used separation process, involves two immiscible phases in which the solutes are distributed according to their affinity. Provided it is fast enough, it allows specific observation of hyperpolarized solutes in each phase, as shown in a preliminary study for a chloroform/water system. The aim of this thesis project is to develop this approach combining hyperpolarization by parahydrogen and extraction, to extend it to new biphasic systems and to apply it to the detection, identification and quantification of very dilute solutes. The ultimate goal is to apply this methodology to samples that contain in essence many solutes, such as those from synthetic chemistry or metabolomics.