06/2014-02/2015, fourth-year undergraduate, Osaka University, supervised by Hidetoshi Shimodaira and me. At the eighth Japanese-French Frontiers of Science symposium (JFFoS), I had the chance to meet Hidetoshi, a famous statistician, who suggested that I co-supervise a fourth-year internship (in Japan, the bachelor lasts four years). With Takamasa, we looked at the Hessian matrix used in waveform inversion in order to find a better way to precondition the inversion. From a statistician’s point of view, the rank of the matrix is extremely low in seismology, which destabilises the inverse problem. Even if this was not a new observation, this report is unforgettable, because since then I pay much more attention whenever I propose to invert seismological data.