Many fields of the industry have recourse to Selective Laser Melting (SLM) to design products that underlie complex geometries. All the stages of the manufacturing process must pass stringent quality controls to verify that the corresponding specifications are well respected. Numerical models help assess the influence of the different machinery settings during the building process, but their simulations are time demanding. To overcome this computational burden, one may rely on Model Order Reduction (MOR), a well-established method that alleviates these specific issues. In the present work, two types of parametric MOR techniques are applied to the case of an SLM heat transfer simulation. The first approach, called intrusive, uses the Galerkin Reduced basis method to solve the system of equations in a much smaller dimensional subspace. An additional layer of approximation, relying on the Energy Conserving Sampling and Weighting[1], is implemented on top of it for more computational efficiency. A second approach, non-intrusive, relies on deep learning [2] to reduce and predict the non-linear time-dependent parametrized parabolic partial differential equation. The authors acknowledge SIM and VLAIO (Flanders, Belgium) for funding the ‘PROCSIMA’ project (M3 research program)