Currently, AM processes such as LPBF are performed open loop, using a fixed, preprogrammed definition of material deposition (path, speed, laser power, and so on). Actual layer and part formation details, even when measured, are not fed back to the print controller to account for actual, as-made layer conditions. Unanticipated layer and part deviations occur frequently that, in the worst case, can result in print failure, part rejection, higher scrap rate, lower yield, and more expensive parts. Process anomalies are sometimes detected manually by the operator. In-process inspection methods such as melt pool monitoring typically do not provide accept/reject guidance. When anomalies are noted, no instructions are provided to the operator or the machine to repair or compensate for the flaw and to salvage the build, in cases where this is possible. We demonstrate development of a closed-loop control capability using a nonthermal in-process inspection method on every layer. Layer Topographic Mapping (LTM) is an in-process inspection method using an optical profilometer to generate a dense, precise map of layer surface height. Algorithms process this data to detect melt flaws with excellent performance. Demonstrated detection of lack of fusion flaws in more than 1,800 Inconel 625 layers is 98.2% probability of detection (POD) and 1.0% probability of false detection (POFD). Optimum repair/rework processes were developed for lack of fusion flaw regions one to three layers thick. LTM software was modified to not only detect flaws but also to define the optimum repair process to employ upon detection based on the number of flaw layers present. Intentionally created (or seeded) lack of fusion flaws were restored to less than 0.1% porosity for one- to three-layer flaws. Porosity in the flaw regions was reduced by up to 98% as verified by CT scans.