Powder bed fusion with laser beam (PBF-LB) is a promising additive manufacturing technique that enables the production of complex geometries with fine resolution and material efficiency, offering significant design freedom and material versatility. However, its broader adoption is limited by the need for extensive parameter tuning, which is often dependent on the specific machine, as well as the material and batch of powder used. In this paper, we introduce a novel algorithm that autonomously identifies melting regimes in an unsupervised manner using optical data acquired from photodiodes — specifically optical emission and reflection. This method eliminates the need for labeled data and achieves an F1-score of 89.2% across both materials tested: Ti–6Al–4V and 316L. Additionally, we propose an uncertainty-driven iterative strategy designed to efficiently generate processing maps by performing experiments based on uncertainty. This approach enables up to a 67% reduction in the number of required experiments, significantly lowering the associated costs of parameter exploration, while sustaining a maximum performance reduction of only 8.88% compared to traditional full factorial designs. Our results demonstrate the potential of this method to streamline PBF-LB optimization, making it more feasible for industrial applications and paving the way for its broader adoption.