A modified supercritical antisolvent method, aiming at high yielding and high loading, is developed for preparation of drug-loaded polymeric nanoparticles. The modified method consists of three alternate stages: injecting solution via a nozzle into a precipitation vessel filled with supercritical CO2, static agitation, and washing of precipitated particles with supercritical CO2. The process stages are continuously repeated until ending injection of a desired amount of solution. The ultrasound operates during the entire process. Poly(lactic-co-glycolic acid) (PLGA) and curcumin are model components. The influences of the process parameters on the yielding and loading are investigated. The maximum yield of the curcumin-PLGA nanoparticles 50 nm in size and the curcumin loading in it reach 96% and 45%, respectively. High yield, high loading, and uniform size are attributed to the efficient mixing between solution droplets and antisolvent, and sequential washing using an antisolvent under ultrasonic agitation. Experimental data are fitted and optimized by neural network simulation.