Recent advances in quantum computing promise computational speedups in several domains including computational engineering. However, current noisy intermediate-scale quantum (NISQ) devices are strongly affected by noise and short decoherence times, which limits their practical reliability [1]. Variational quantum algorithms partially mitigate these issues, but are typically developed and tuned on quantum circuit simulators. State-of-the-art noisy simulators rely on density-matrix representations, whose memory footprint scales as 𝒪(4ⁿ) for an n-qubit system, restricting simulations to comparatively small problem settings, often with about half the qubits accessible compared to statevector simulators. In this work, we follow a different approach and present a Python package that augments a statevector simulator with hardware-specific noise models derived from quantum device calibration data, enabling efficient approximation of density-matrix simulations. The package supports automatic noise model generation, transpilation of selected quantum gates to the basis gate set and qubit layout of a target backend, and submission of quantum circuits to real hardware for execution. We benchmark our simulator against state-of-the-art density-matrix backends such as PennyLane [2], evaluating output-state fidelity, runtime, and memory consumption. The results show that our approach closely reproduces the density-matrix outputs while achieving substantial runtime improvements and significantly lower memory usage, especially at higher qubit counts. Finally, we demonstrate the package on an engineering application by training a quantum neural network with the noisy simulator and validating it on quantum hardware. The network predictions obtained from simulation and hardware exhibit close agreement, with less than 0.2% deviation in the R² score, illustrating both the feasibility and accuracy of the proposed framework for NISQ-era quantum machine-learning experiments.