Thermal conductivity of shallow (<5m) soil is a critical property for the design of subsea cables and pipelines. In complex geological settings, thermal conductivity can vary greatly both with depth and along the cable or pipeline route, and the standard laboratory approach of discrete needle probe testing can fail to characterise thin layers or gradual changes. In this paper, continuous depth profiles of thermal conductivity are predicted from Multi-Sensor Core Logging (MSCL), a non-destructive, high-resolution (cm-scale) method to measure soil properties on recovered samples. A porosity-thermal conductivity relationship is derived and is well approximated with the weighted geometric mean equation, with the coefficient of determination, r2, = 0.77 and root-mean squared error (RMSE) = 0.3 W/mK. Furthermore, bulk density and natural gamma data from the MSCL is used to automatically classify soil samples into three categories: clay, sand, and organic soils. Soil-specific relationships between porosity and thermal conductivity improve the prediction of thermal conductivity with r2 = 0.84 and RMSE = 0.25 W/mK. This study highlights the ability to predict thermal conductivity and soil type from MSCL data, and the implication that including MSCL in a laboratory program can reduce the total volume of destructive testing required.