Numerical simulation of turbulent combustion requires reduced-order models that capture multi-scale physics at affordable computational cost.[1] For non-premixed flames, the flamelet approach represents a turbulent flame as an ensemble of laminar diffusion flamelets and couples CFD to an a priori tabulated library of solutions of the transient flamelet equation.[2] While accurate, tabulation introduces two persistent bottlenecks: (i) large memory footprints to keep discretization errors small and (ii) an interpolation step that adds runtime and can introduce errors when the library is sampled coarsely. We propose a physics-informed neural network (PINN)[3] that acts as an ML-enhanced PDE solver for the transient flamelet equation, a stiff reaction-diffusion system. To handle strong diffusion-reaction coupling and chemical stiffness, we embed a differentiable chemistry integrator into the physics-based loss, enabling backpropagation through the full reaction mechanism while enforcing the governing PDE together with boundary and initial constraints. The model is trained in forward mode and conditioned on strain rate, providing a single surrogate over a practically relevant range. Validation on hydrogen diffusion flamelets, an ideal stress test due to pronounced kinetic stiffness, shows that the PINN reproduces reference flamelet solutions with high fidelity for individual strain rates and maintains accuracy when generalized across strain rates. By replacing discrete tables with a compact, continuously differentiable surrogate, the approach reduces the tabulation memory footprint by 370.8 and enables a speed-up in the computation of flamelet data over multiple strain rates compared to classical numerical solution by a factor of 10.7. Furthermore, the use of a PINN removes interpolation artifacts and supports scalable CFD coupling, contributing to ML-enhanced PDE solvers for combustion manifolds in computational science. [1] Posch, S., Gößnitzer, C., Lang, M., Novella, R., Steiner, H., Wimmer, A., Turbulent combustion modeling for internal combustion engine CFD: A review, Progress in Energy and Combustion Science, 106, 101200, 2025. [2] Peters N., Turbulent Combustion, Cambridge Monographs on Mechanics, 2004. [3] Raissi, M., Perdikaris, P., Karniadakis, G. E., Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, Journal of Computational physics, 37