<p>Unlike traditional 3D printing on flat planes, nonplanar printing creates true 3D curved layers, boosting strength, surface quality, and reducing support needs. Achieving optimal layers requires a slicing algorithm that accounts for geometry and load conditions. This paper presents the Multi-step Evaluation of Transitioned Anisotropy (META) Slicer, a novel optimization-based nonplanar slicing method that incorporates real-time updates of stress distributions into the slicing process to maximize mechanical strength in anisotropic 3D printed parts. META Slicer builds on the Neural Slicer framework by integrating Finite Element Analysis (FEA) updates into the layer deformation process, aligning layers with the direction of maximum principal stress (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\:{\widehat{\tau\:}}_{max}\)</EquationSource> </InlineEquation>). To overcome the computational cost of iterative FEA during slicing, a deep neural network, named Stress Flow Net (SFN), is trained for each case to rapidly predict <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\:{\tau\:}_{max}\)</EquationSource> </InlineEquation> directions based on part geometry and material orientation, making the process up to 100 times faster and feasible. META Slicer is validated through numerical simulations on five case studies and experimentally tested on two of them using a 6-axis robotic 3D printing framework. Results demonstrate up to 31% reduction in von Mises stress and 71% reduction in maximum principal strain compared to planar slicing, and outperform Neural Slicer in most cases. These findings confirm META Slicer’s potential as a transformative tool for strength-aware 3D printing of anisotropic materials and its applicability to a wide range of future engineering scenarios.</p>