Noise remains the central obstacle to extracting reliable computational results from quantum hardware. This proposal outlines a unified research program targeting hardware-efficient approaches to noise suppression and error correction across two complementary paradigms of quantum computing. For analog quantum simulation tasks, we develop a novel energy-gap-protection scheme against 1-local coherent noise, utilizing an excited encoding subspace stabilized by solely 2-local commuting Hamiltonians. This approach bypasses existing no-go theorems and scales polynomially with system size. In the digital quantum error correction (QEC) domain, we investigate belief-propagation-based decoding algorithms that can meet the stringent ~1μs latency budget of superconducting qubit architectures, with the goal of building an end-to-end automated pipeline for designing, training, and deploying real-time QEC decoders. The proposed research bridges these two directions by exploring measurement-based feedback as a mechanism to extend error suppression from coherent to incoherent noise, and by studying real-time decoding of circuits with nontrivial logical operations. Together, these contributions aim to make quantum computing more reliable and practical on near-term hardware.