What? Surely deployment is by far the easiest part of machine learning? The learning is the hard part. If you are resource constrained, then learning a small/efficient model is even harder, obviously. As for binary math, usually all you need is the standard bitwise operators and popcount.