Halos is a new system-level tester under development, with an automation sub-system that is a gantry style robot. The gantry beam has four sets of pick/place units running along its linear track. Each pick/place unit itself is a Cartesian style robot with its own X-, Y-, Z-, and R-axes. Therefore, a row of four devices under test (DUT) can be picked or placed simultaneously. This presentation discusses a series of accuracy related challenges and demonstrates how to calibrate the system to achieve required placement accuracy. The system relies on vision as its sensor of choice, with cameras mounted on the pickers, facing down, measuring the pick/place sites, and cameras mounted below the pickers, facing up, measuring the DUT offsets while holding the nozzle vacuum. The solution starts with calibrating the cameras, then proceeds to a number accuracy related challenges that will be detailed in the presentation. Only then can hand-eye coordination be achieved: namely the picker nozzle can accurately reach the camera-measured placement site. This presentation describes how we model the sources of inaccuracy and the techniques to calibrate them out.