Efficient thermal management has become increasingly important as modern electronic devices continue to shrink in size while operating at higher power densities, leading to elevated heat fluxes and localized overheating. Polymer composites offer an attractive platform for thermal management due to their low density, processability, and cost-effectiveness; however, their inherently low thermal conductivity limits their performance. Incorporation of graphene- and graphite-based fillers has emerged as a promising strategy to enhance heat dissipation; however, achieving high thermal conductivity in polymer composites remains challenging due to interfacial thermal resistance, filler agglomeration, and degradation of graphene’s lattice during chemical modification. Consequently, optimizing both polymer–filler interfacial interactions and the preservation of intrinsic phonon transport pathways is essential. In this work, the influence of graphene functionalization strategies and synthesis routes on thermal transport was systematically investigated in polyetherimide (PEI) composites. Several graphene-based fillers, including pristine graphene nanoplatelets (GnPs) and thermally expanded graphite (EG), were examined to elucidate how chemical modification and structural integrity govern heat conduction in polymer matrices. Edge-oxidized graphene (EGO) was prepared using a controlled edge-selective oxidation approach that preserved the sp²-bonded basal plane while introducing oxygen-containing functional groups primarily at the sheet edges. This selective functionalization enhanced polymer–filler interfacial bonding without significantly disrupting phonon transport, resulting in an 18% increase in thermal conductivity at 10 wt% filler loading compared to pristine GnP/PEI composites. In contrast, basal-plane-oxidized graphene (BGO), produced using the Hummers method, introduced oxygen functionalities directly onto the basal plane, leading to increased defect density and lattice distortion. These structural disruptions significantly impeded phonon propagation, causing a 57% reduction in thermal conductivity relative to pristine graphene-filled composites. Raman mapping, X-ray photoelectron spectroscopy (XPS), and Fourier transform infrared spectroscopy (FTIR) confirmed preferential edge functionalization in EGO and extensive basal-plane disorder in BGO, underscoring the importance of oxidation site control. To further enhance long-range phonon transport, EG was incorporated into PEI via a solvent-casting technique. EG, obtained by thermal expansion of graphite intercalation compounds, forms a three-dimensional interconnected network of continuous graphene sheets with minimal interlayer contact resistance. Preservation of the worm-like porous EG structure during solvent casting enabled the formation of an interpenetrating polymer–graphite network conducive to efficient phonon transport. As a result, thermal conductivity increased nearly linearly with EG loading, reaching 7.3 W/(mK) at 10 wt%, a substantial enhancement compared to pristine PEI (≈0.23 W/(mK)). The effect of intercalation chemistry on EG microstructure and performance was further evaluated using two expansion routes: H₂SO₄/H₂O₂ and H₂SO₄/NaClO₃. EG synthesized via the H₂O₂ route exhibited superior structural integrity, predominantly edge-localized oxidation, and enhanced graphitic continuity, resulting in thermal conductivities as high as 9.5 W/(mK) at 10 wt%. In contrast, EG produced using NaClO₃ showed greater basal-plane damage and reduced network continuity, yielding a lower thermal conductivity of 5.3 W/(mK). Raman ID/IG ratios, XPS oxygen speciation, and X-ray diffraction analysis corroborated these structural differences. Overall, this study demonstrates that controlling oxidation location, minimizing basal-plane damage, and preserving long-range graphitic networks are critical for achieving high thermal conductivity in polymer–graphene composites, providing key design principles for scalable thermal interface materials and advanced heat-spreading technologies.