Human serum albumin (HSA) nanoparticles are promising drug carriers; however, robust control of their physicochemical and biological properties remains challenging. Here, we investigate how formulation parameters, mixing regime and crosslinking chemistry govern HSA nanoparticles' biological performance. Glutaraldehyde (GA)-crosslinked HSA nanoparticles were first optimised using a bulk desolvation process by varying HSA concentration, pH and desolvation solvent. The formulation was then translated to a staggered herringbone micromixer to tune particle size via total flow rate and flow rate ratio. In parallel, a crosslinker-free route based on glutathione (GSH)-mediated disulfide reorganization, with or without H2O2, was developed, in which GSH transiently reduces intramolecular disulfide bonds and subsequent oxidation forms stabilising intermolecular disulfides. GA-crosslinked and GSH-mediated disulfide-crosslinked HSA nanoparticles were compared in terms of colloidal and serum stability, cytotoxicity and cellular uptake in 4T1 breast cancer cells, and curcumin was encapsulated as a model hydrophobic drug. Both strategies produced physically stable nanoparticles, curcumin-loaded nanoparticles with good drug loading. In carrier-focused evaluations, GA-crosslinked HSA nanoparticles exhibited smaller sizes, negative charge, and high biocompatibility, with sustained cellular uptake and minimal serum-induced aggregation. In contrast, GSH-based HSA NPs were strongly cationic, exhibited faster initial cellular uptake, and underwent pronounced serum-mediated charge reversal, under the tested conditions. Using curcumin (CCM) as a model hydrophobic payload, both formulations achieved effective drug encapsulation (GA-crosslinked 17.89% and GSH-based 14.6%). GA-crosslinked CCM-HSA NPs showed a small particle size, low PdI, and demonstrated enhanced antiproliferative activity compared with free curcumin. These results provide design guidelines for engineering albumin-based nanocarriers with tailored stability and drug delivery performance.