The mathematical comprehension of biological tissues remains a significant challenge, where Finite Element Method (FEM) simulations [1, 2] have emerged as essential tools for understanding human tissue behaviour. However, generating patient-specific meshes from CT or MRI scans is often a time-consuming bottleneck [3]. This work presents a preliminary study on a neural network designed to automate the meshing process of the Achilles tendon (AT) [4]. By inputting key morphological parameters, such as total length, width at extremities and mid-section, and torsion angle, the model aims to output valid, simulation-ready meshes. As observed by Pękala et al. [5], the AT is formed by three subtendons and classified into three types, depending on the degree of torsion between the proximal end and the insertion with the calcaneal tuberosity. This study aims to develop a neural network that reconstructs the AT considering this classification, ensuring the three subtendons are distinct for FEM analysis, following the Knaus and Blemker methodology [6]. This integration of deep learning and clinical imaging enables near real-time generation of patient-specific models, potentially enhancing therapeutic interventions and expediting recovery processes in tendon degeneration. The methodology comprises three stages: first, a 3D Achilles tendon model is reconstructed from MRI scans and refined in ANSA to generate a high-quality tetrahedral mesh. Second, a stochastic algorithm in MATLAB generates 1,000 synthetic samples through controlled morphometric variations to create a robust training dataset. Finally, a neural network is trained to predict nodal coordinates based on the input parameters. This framework bypasses traditional pre-processing constraints, providing a fast, reliable tool for advanced biomechanical analysis.