By tying the knowledge to the company itself. Knowledge of a proprietary control system, its configuration conventions and its failure modes is knowledge of the petitioner's product; general controls or millwright competence is a trade skill. The evidence is internal: documents the employee produced, training the company gave, and what replacing him would take.
Test the claim against the labour market
Ask a simple question and answer it honestly in the petition: could the subsidiary hire someone in the United States to do this next month? If the answer is that a qualified controls technician could be hired but would need four to six months of training on the company's own system before commissioning a machine unsupervised, that is the argument, and it should be stated with the reasoning behind the estimate. If the answer is that anyone with the relevant trade ticket could do the work, the case is weak and a different approach is needed. Support the answer with the internal record: the versions of the configuration library, the dates the fault routines were written, the machines commissioned and by whom. Facts of that kind survive scrutiny in a way that superlatives do not.
Hypothetical example: a laser-guided lumber-sorting engineer designed the error-correction tables used only in her employer’s mills. A first review collects version histories, training time for new technicians, access controls and records of installations she corrected. These records settle whether the knowledge is particular to the organization rather than advanced trade skill. If an outside engineer could perform the U.S. job after ordinary orientation, the case needs reconsideration. Confirm separately that she completed the required uninterrupted foreign year with the qualifying organization in the three years before filing.