The paper tackles a key practical challenge in bringing quantum key distribution (QKD) to resource-constrained devices: the computational and memory overhead of classical post-processing, particularly error correction. The researchers develop a composable finite-size security framework with a tight bound on error-correction leakage using non-binary low-density parity-check (LDPC) codes. Crucially, the framework connects achievable secret-key rates with the underlying code parameters and the memory required for error-correction encoding. The study also demonstrates how sparse representations of parity-check matrices can significantly reduce storage requirements, providing a practical way to assess the feasibility of continuous-variable QKD (CV-QKD) on devices with limited resources. The work also explores an asymmetric implementation model in which a lightweight device performs the comparatively less demanding encoding operation, while a more capable receiver or gateway carries out the computationally intensive decoding. Such an architecture could help enable short-range quantum-secure communications for future IoT devices, sensors, wearables and drones.