In Antarctica, ice crystals (platelets) form and grow in supercooled waters below ice shelves. These platelets rise, accumulate beneath nearby sea ice, and form a several meter thick sub-ice platelet layer. This special ice type is a unique habitat, influences sea-ice mass and energy balance, and its volume can be interpreted as an indicator for ice – ocean interactions. Although progress has been made in determining and understanding its spatio-temporal variability based on point measurements, an investigation of this phenomenon on a larger scale remains a challenge due to logistical constraints and a lack of suitable methodology. In the present study, we applied a lateral constrained Marquardt-Levenberg inversion to a unique multi-frequency electromagnetic (EM) induction sounding dataset obtained on the ice-shelf influenced fast-ice regime of Atka Bay, eastern Weddell Sea. We adapted the inversion algorithm to incorporate a sensor specific signal bias, and confirmed the reliability of the algorithm by performing a sensitivity study using synthetic data. We inverted the field data for sea-ice and sub-ice platelet-layer thickness and electrical conductivity, and calculated ice-volume fractions using Archie’s Law. The thickness results agreed well with drillhole validation datasets within the uncertainty range, and the ice-volume fraction also yielded plausible results. Our findings imply that multi-frequency EM induction sounding is a suitable approach to efficiently map sea-ice and platelet-layer properties. A successful application of this technique requires a break with traditional EM sensor calibration strategies due to the need of absolute calibration with respect to a physical forward model.