Direct laser interference patterning (DLIP) is a well-established technique for fabricating micro- and nano-scale structures that can enhance the properties of surfaces such as reduced friction and wear. However, achieving full automation requires reliable in-line process monitoring to ensure consistent structure quality. In this study, an infrared monitoring camera is implemented to capture spatially resolved temperature distributions during DLIP processing. Stainless-steel samples are structured while systematically varying the laser fluence (2.5–5.6 J cm⁻²), and path velocity (1–20 mm s⁻¹). The resulting surface structures are characterized using confocal microscopy to extract key topographical parameters. A convolutional neural network is trained using 180 000 process images from the IR system and the corresponding topographical data. The model identifies clear correlations between laser fluence, thermal signatures, and surface topography. For specific parameters, prediction accuracies of up to 94% are achieved. These results demonstrate that combining infrared monitoring with machine learning enables indirect yet accurate prediction of surface features, paving the way for enhanced process control and quality assurance in DLIP and related manufacturing processes.