Protein detection plays an essential role in early disease diagnosis and biodefense applications. One commonly used clinical technique for protein detection is the Enzyme Linked Immunosorbent Sandwich Assay (ELISA) which utilizes antibodies to capture targets to produce colorimetric signal. However, signals generated from unaccounted non-specific binding by antibodies result in high false-positive readouts, leading to misdiagnosis in patients. To address these issues, we use Surface Enhanced Raman Spectroscopy (SERS) together with the ELISA sandwich model to create a more accurate assay. Recently, SERS has been gaining momentum as a highly sensitive technique for detecting biomolecules in complex media. Using the model protein tumor necrosis factor-alpha (TNF-α), a protein with fundamental functions in inflammation and immune system development, we created an assay that eliminates non-specific binding while also reporting a precise readout of true protein binding. In this assay we immobilized two distinct affinity reagents specific to TNF- α onto gold nanoparticles (AuNP) and gold film pre-labelled with Raman reporters. TNF- α mediates the joining of the AuNP with the film to produce surface enhanced Raman signals that correlate the concentration of the protein to the intensity of the signal. A series of passivating agents were also considered to decrease non-specific binding. With this construct, standard curves were produced in buffer solution with a limit of detection around 100 pM, well within the range of TNF- α found in the body. This assay demonstrates the possibility of monitoring disease states through quantification of key biomarkers.