Artefacts generated by the proprietary work: design documents, release notes, revision histories naming the specialist, internal training records, and customer contracts describing delivered systems. Also a time allocation showing what proportion of the assignment is proprietary work, kept as it accrues.
Evidence produced while the work happens
An extension file built from contemporaneous artefacts is dramatically stronger than one assembled retrospectively from job descriptions. The most useful items come out of ordinary work. Design and architecture documents for the integration layer, with authorship visible.
Release notes and revision histories that show him making the changes. Internal training records showing which local staff he has qualified on the company's own procedures, which simultaneously demonstrates the knowledge is proprietary and that he is transferring it. Customer contracts and statements of work that describe a system being delivered by the petitioner, supporting the position where work is performed on a client's premises.
And a simple record of how the assignment's time is allocated, kept quarterly. None of that is created for the immigration file; it is created by doing the work properly, which is why it persuades. Hypothetical example: a semiconductor test-equipment maker needs the engineer who designed its calibration sequence at its Nevada affiliate.
The first review produces a knowledge inventory tied to dated training, access controls, project records, and the proposed duties. Technical manuals, version histories, and supervisor letters settle what is company-specific and how the engineer gained it; payroll settles the qualifying foreign employment. The company should secure copies before restricted systems change or colleagues leave.
A bare résumé proves ability, not specialised organisational knowledge, and a generic service job description can contradict otherwise strong technical evidence.