This paper introduces a novel framework that fuses symbolic innovation heuristics with explorative generative artificial intelsligence for early-stage inventive design. Grounded in the hypothesis that symbolic reasoning and large language models (LLMs) are complementary, the method operationalizes TRIZ and SAVE principles as structured prompts to guide generative exploration. The design process is framed as a semantic traversal of a latent invention space, where symbolic heuristics act as attractors shaping the outputs of the LLM. Instead of applying fixed rules, the framework enables a dynamic interaction between symbolic intent and generative flow, producing design hypotheses that emerge through inference rather than enumeration. Architecture supports both systemic and component-level innovation, fostering divergent solution paths that converge into coherent design proposals. A case study on an autonomous drone with hybrid vertical and fixed-wing flight demonstrates the framework’s ability to generate novel, principle-aligned concepts. The results suggest broad applicability to complex systems and future potential for integration with real-time evaluation tools. The proposed approach marks a step toward AI-assisted invention that is both structured and creatively expansive.