Laser-based techniques such as Direct Laser Interference Patterning (DLIP) have significantly impacted surface functionalization, enabling the fabrication of a wide variety of periodic structures with complex geometries. [1] Real-time sensing methods are crucial to ensure surface quality and enable effective process monitoring, especially for industrial applications. [2][3] In this study, a set of signals obtained from a photodiode (nLIGHT, Plasmo) integrated into a picosecond-pulsed DLIP system (ELIPSYS®, SurFunction GmbH, see Figure 1a) are analysed to predict surface roughness features (Sa, Sq, Depth, etc.) on microstructures produced on aluminium AA2024. The investigation involved varying DLIP process parameters (repetition rate and laser power), which resulted in different surface roughness values. The real-time recorded signals (Figure 1b) were analyzed using machine learning algorithms (such as Random Forest, see Figure 1c). The results show that laser parameters and surface characteristics in DLIP can be estimated from the sensor signal, thereby providing the basis for real-time monitoring in DLIP structuring processes.