The ordering in the target list is intentional. Engine, platform, and field teams hire against exactly the artifacts this roadmap produced: verified endpoints, benchmark posts, merged PRs, a cost model from your own fleet. Kernel-specialist roles want a deeper kernel portfolio, so they come later, once those artifacts are strong. Apply in batches rather than serially; loops run for weeks, and parallel processes are offer leverage at the end.
Treat the first loop as a calibration exercise you expect to fumble. Write down every question you couldn't nail and feed it back into the gauntlet before the loops you actually care about. That feedback cycle, not any single prep session, is what converts the work into the title.
Archetype B/C/E first: engine teams (Together, Fireworks, Baseten, Red Hat/vLLM), platform teams (Anthropic, OpenAI, CoreWeave, Modal), field roles. Kernel roles (archetype A) once Phase 4 artifacts are strong.
build+150 XPThe first loop is a calibration exercise. Take notes on every question you couldn't nail; feed them back into the gauntlet.
build+200 XPThe final boss. (Comp data point from the research: senior inference roles at specialist shops clear $400k+ TC.)
build+1000 XP