Quantum sensors based on the interference of cold atoms have advanced to the forefront of precision measurements in geodesy, metrology and tests of fundamental physics. The ultimate potential of these devices is realized using quantum degenerate atoms in extended free fall. This can be achieved on microgravity platforms such as drop towers, parabolic flights, ballistic rockets, satellites and space stations. The transition to mobile and robust devices that can withstand the demands of these environments comes with many challenges. Quantum sensors need to be scaled down and integrated without compromising their performance. In fact, they need to significantly outpace conventional instruments, since microgravity time is an expensive resource and limited to a few seconds at a time on the most accessible platforms. This thesis describes the construction, qualification and operation of a miniaturized ultracold atom experiment that meets these challenges. The QUANTUS-2 apparatus features a payload weight of 147 kg and a payload volume of 0.3 m^3. It generates Bose-Einstein condensates of 4×10^5 ^87Rb atoms every 1.6 seconds, a flux of ultra-cold atoms that is on par with the best lab-sized devices. Ensembles of 1×10^5 atoms can be created at a 1 Hz rate. It is currently the fastest machine of its kind and achieves the highest atom number of any atom chip setup. The apparatus continuously withstands peak accelerations of up to 45 g during microgravity campaigns at the drop tower facility in Bremen, Germany. Here, the payload has accrued 208 drops and 9 catapult launches over 24 month. The setup is the first atom optics experiment to stand up to the technical demands of catapult operation. Four condensates can be created and observed consecutively during nine seconds of free fall in a single catapult launch. In total, the experiment has been suspended in microgravity for over 17 minutes. With the record source performance, the repetition rate for microgravity experiments with ultra-cold atoms was increased by a factor of four compared to previous devices. The total atom number was increased by a factor of 40, vastly improving the signal to noise ratio for absorption images of spatially extended clouds. The ensembles can be prepared consistently over many weeks of drop tower operation. The variance of the mean center of mass velocity in two observable directions is 7.3 μm/s and 6.9 μm/s. Magnetic lensing techniques were employed to manipulate the expansion of the ensembles. First results yield a residual expansion rate in three dimensions of σ_v = 116.9 ± 13.9 μm/s, which implies a three-dimensional effective temperature of T = 47.6 ± 11.3 pK at an average condensate atom number of N = 93000. These values constitute the best collimation of any atomic ensemble and the most promising source for atom interferometry reported to date. Optimizing the current lensing sequence will reduce the expansion rate further to effective temperatures in the femtokelvin regime. The level of control demonstrated over the condensates is highly relevant for the advancement of matter-wave optics and quantum sensors. Controlling the motion and size of atomic clouds is intrinsically tied to many systematic effects in high precision measurements. QUANTUS-2 will provide a platform to explore and mitigate these limitations on unprecedented time scales of up to seven seconds of free evolution.