Additive Manufacturing (AM) has greatly expanded the design possibilities for engineering components by enabling the fabrication of highly complex geometries. This has stimulated growing interest in lattice structures as advanced engineering materials, as they allow lightweight designs with tunable mechanical performance. A major advantage of lattice architectures lies in the ability to tailor both material distribution and unit cell (UC) geometry to meet specific functional requirements, enabling the optimization of mechanical properties and extending their applicability across multiple industrial sectors. Despite these benefits, heterogeneous lattice structures—featuring spatially varying microarchitectures within a single component—remain relatively underexplored. Their limited adoption is mainly due to the challenges associated with ensuring geometric continuity and structural integrity when connecting unit cells with different topologies. Although alternative heterogeneous infill strategies have been proposed, they often prioritize self-supporting configurations to facilitate manufacturability, which can lead to reduced performance under non-uniform or complex loading conditions. In previous work, the authors introduced a geometric algorithm for generating beam-based unit cells with variable topology while guaranteeing full connectivity across heterogeneous lattices, making them suitable for AM. Building upon this approach, the present work proposes an optimization framework for components designed using this methodology. The lattice structures are represented as equivalent continua with spatially varying effective material properties, expressed as smooth functions of the unit cell geometry. This formulation enables reduced-order modeling of the lattice microstructure, significantly lowering computational costs while maintaining accuracy. To further simplify the optimization problem, effective material properties are assumed constant within each unit cell, reducing the dimensionality of the design space, mitigating over-parameterization, and improving numerical robustness and convergence. The proposed workflow integrates CAD modeling through OpenCascade, adaptive mesh generation via Gmsh, and finite element analysis using FEniCSx, providing a streamlined design-to-simulation pipeline. Finally, a demonstrative component was fabricated and validated through numerical and experimental testing.