Conventional public-key cryptosystems such as RSA and standard elliptic curve cryptography (ECC), although widely deployed, remain vulnerable to quantum adversaries and incur prohibitive computational overhead when applied to microcontroller-class IoT endpoints. This paper presents a novel hybrid cryptographic framework that integrates Ntru-lattice-based encryption (NTRU) with elliptic curve Diffie-Hellman (ECDH) key exchange and SHA-3 integrity verification to construct a layered, post-quantum-resilient security architecture. The proposed scheme exploits the algebraic structure of polynomial rings over truncated polynomial quotients to achieve both hardness guarantees rooted in the shortest vector problem (SVP) and the computational efficiency required by embedded platforms. We formalize the security reduction under the ring learning with errors (RLWE) assumption and provide a rigorous mathematical analysis of key generation, encryption, and decryption sub-protocols. Experimental evaluation on ARM Cortex-M4 and RISC-V FPGA platforms demonstrates that the proposed scheme achieves 256-bit equivalent security with encryption latency of 9.7 ms, outperforming RSA-2048 by 78% and ECC-256 by 47% in identical deployment configurations. Memory footprint is reduced to 3.2 KB compared to 11.4 KB for conventional RSA implementations.