Growing consumer interest in high protein foods and beverages has fueled market demands for milk-based beverages with high dairy protein content. To capture this market opportunity, dairy processors need to produce dairy-based beverages that remain physically stable and free of off-flavors until the end of shelf life. Milk-based beverages with higher protein content are more susceptible to loss of physical stability by gelation or phase separation. An understanding of the impact of different dairy cow genetics, cow feeding, product formulation, processing, and storage conditions on the protein fraction of milk-based beverages is needed to develop milk-based beverages that remain acceptable to consumers throughout shelf life. The effects of protein concentration, thermal treatment, storage time, and storage temperature on the protein fraction of milk-based beverages were investigated. Liquid micellar casein concentrates (MCC) and milk protein concentrates (MPC) are produced by the microfiltration and ultrafiltration of skim milk, respectively. These concentrates can be used as the base ingredients for producing milk-based beverages. The mouthfeel and viscosity of liquid MCC-based beverages depend on the beverage protein content as well as the composition of soluble proteins in the aqueous phase around the casein micelle (supernatant). As protein content of the MCC-based beverages increased, the concentration of supernatant protein and apparent viscosity of the beverage also increased. Predicted MCC apparent viscosity was also positively associated with the dissociation of αs- and β-caseins, which increased as temperature decreased due to weakening hydrophobic interactions. Optimal beverage viscosity could be achieved by controlling the dissociation of these proteins in MCC. MCC and MPC-based beverages with and without 0.15% dipotassium phosphate (DKP) were process stable even at ultra-high temperature (UHT) treatment using direct steam injection. DKP addition increased the concentration of supernatant protein, phosphorus, and calcium concentrations, especially in MCC-based beverages. These concentrations increased over storage time, especially when lower heat treatments (high temperature short time [HTST] or no heat treatment) had been applied. DKP could be removed from beverages made with fresh liquid MCC and MPC (containing a residual lactose concentration of 0.6 to 0.7% and the proportional amount of soluble milk minerals). Age gelation occurred in both commercial aseptic milks (thermally treated by direct steam injection) stored at 4 and 21oC for 12 months. During storage, little to no casein proteolysis products formed, indicating that non-proteolytic changes in the milk were responsible for age gelation of the aseptic milks. The concentration of the highly hydrophilic κ-casein-whey protein complex increased in concentration in the liquid phase of the milk, while the gel phase was enriched in the more hydrophobic αs-casein and β-caseins. Age gelation was driven by dissociation of κ-casein out of the casein micelles due to heat-induced disulfide binding of whey proteins to κ-casein during UHT and lactosylation of the bound whey proteins via Maillard reaction over storage time. Milk-based beverages may develop off-flavors and lose physical stability through enzymatic damage by proteolysis and lipolysis. Proteolysis produces bitter peptides and reduces casein micelle stability, while lipolysis produces rancid off flavors. Prevention of enzymatic damage to milk protein and fat first requires a method for monitoring and measuring levels of proteolysis and lipolysis activity. Partial least squares (PLS) prediction models for rapidly measuring proteolysis (i.e., concentrations of total casein proteolysis products [CNPP] and an individual CNPP with a molecular weight of approximately 12 kDa) and lipolysis (i.e., concentrations of total free fatty acids [FFA] and short chain FFA) in milk using mid-infrared (MIR) spectroscopy were developed. The ratio of performance to deviation (RPD) of the total CNPP and CNPP #6 models were 1.8 and 1.6, respectively. Weak signals from the absorbance of CNPP in milk and extensive overlap of absorbance by other milk components at critical wavenumbers strongly associated with CNPP signals appeared to limit the capability for predicting CNPP concentration from MIR spectra. Separating and concentrating the aqueous phase of the milk where most of the low molecular weight proteolysis products reside may be necessary for improved performance in measuring casein proteolysis using MIR analysis. The RPD of the total FFA and short chain FFA models were 7.0 and 5.9, respectively. The standard error of prediction on validation of the total and short chain FFA models were 2.84 and 0.31 mg/100 g milk, respectively. Based on the RPD values and validation performance, the total and short chain FFA models would be suitable for use as an indirect rapid method for quality assurance monitoring of free fatty acids levels in raw bulk tank milk samples.