The Small Punch Test (SPT) has shown significant potential for the estimation of material properties and mechanical assessment of in-service components in various industries, particularly the nuclear fusion industry. It is considered a simple and economic experimental procedure where relatively very small samples, taken from engineering components, are required to evaluate material properties. The low volume of material required for the SPT makes it desirable for assessing irradiated components, which is essential to maintaining nuclear fusion reactors. That is, the small-scale nature of the test makes it suitable for studying irradiated materials and irradiation damage in fusion reactor plasma-facing components, where the extraction and handling of large volumes of materials required for traditional tensile tests are challenging and require robust management of safety conditions and increases the risk of exposure to radioactivity. However, the main challenge associated with the SPT is the complex interpretation of the test results and their correlation to macroscale tensile properties such as yield stress, ultimate tensile strength and ductility. Due to the challenge in correlating material behaviour across length scales, results are typically analysed to assess relative changes rather than to derive absolute engineering values. Because the absolute values are needed to understand component-scale changes, several methods have been proposed to correlate SPT results with traditional stress-strain data using finite element analysis (FEA) and inverse analysis. However, there has been no generalised and material-independent robust approach for obtaining the desired tensile material properties from SPT results. Most of these methods correlate particular points on the load-displacement curves, such as maximum force, elastic-plastic transition point and plastic instability point, with tensile material properties. However, these correlations are not entirely material-independent, and identifying these points requires carrying out extensive modelling and analysis for each test. This paper applies Gaussian Process Regression (GPR) to reconstruct stress-strain curves from SPT results (load-displacement curves). GPR models have been widely used in Machine learning applications due to their representation flexibility and inherent uncertainty measures over predictions. To achieve this, stress-strain data of a wide range of hypothetical materials with semi-randomised hardening parameters were obtained based on the Voce Law Nonlinear Isotropic Hardening model. These material models were used as input into a Small Punch Test FEA model to obtain the SPT results (load-displacement curves). The synthetic load-displacement curves and resulting material hardening parameters were used to train a multivariate gaussian process regression model. The relationship between the hardening parameters (and hence stress-strain data) of these hypothetical materials and the SPT results (load-displacement curves) was established based on the trained GPR model. The GPR model predictions were validated using SPT results of standard materials of known stress-strain data, namely Stainless Steel 316. While Artificial Neural Networks promise accurate predictions, they require massive simulation data. GPR, on the other hand, is considered less data-hungry and less computationally expensive and offers reliable uncertainty measures. Compared to existing methods, this novel approach offers a faster and material-independent regime to determine tensile properties from Small Punch experiments.