Flow cytometry is widely used for cell identification and characterization and involves labelling biological and clinical samples with fluorochrome-conjugated antibodies specific to cell markers. This requires use of expensive exogenous reagents and necessitates complex pre-processing of samples. Additionally, extensive challenges arise in clinical samples consisting of highly plastic and heterogenous cell types observed in diseases such as cancer. As such, it is challenging to apply flow cytometry to point-of-care diagnostic applications. To address this issue, we investigated the combination of diffuse reflectance spectroscopy (DRS), microfluidics and machine learning to offer rapid, low-cost, label-free cell identification for potential deployment at the point of care. To achieve this, we utilized a compact fibre-optic diffuse reflectance spectrometer with multi-depth sensing capability. This system was applied to a proof-of-concept cell identification study where we were able to discriminate triple negative breast cancer cells from healthy fibroblasts using commercially available flow channel slides (Ibidi GmbH, channel dimensions: 5 mm width, 0.4 mm height). However, we observed high interexperimental variability, which was partially attributed to the relatively large fluidic channels. Thus, we investigated in-house fabrication of microfluidics of varying channel widths (0.6-2mm). To this end, we used a Mars ELEGOO 3D printer and commercially available printing materials to batch fabricate optically and mechanically viable microfluidic chips that were both cheap and customizable. Using these in-house microfluidic devices, we demonstrated DRS-based discrimination of cancer cells of different origins, further indicating the potential of this approach for point-of-care cell identification/characterization. Ultimately, we hope this work will lead to the development of cheap, deployable, and accurate point-of-care tools for rapid, label-free cell identification.