Roadmap

The roadmap starts with evidence, not dates

A public commitment to build toward inference infrastructure engineering by mid-2029, with capacity and evidence gates instead of fixed completion dates.

The roadmap starts with evidence, not dates.

I’m building toward becoming an inference infrastructure engineer by mid-2029: owning serving systems end to end and going one layer deeper into engine internals, without pretending to compete with full-time kernel authors. The route is Kubestronaut—KCNA, KCSA, CKA, CKAD, and CKS all ACTIVE simultaneously, targeted for December 2027—then MLOps and GPU serving, then inference engine internals.

The schedule has a capacity dial: survival is 3 h/wk, normal is 10, and surge is 20. Milestones unlock on evidence rather than dates, so dates float; there is no completion percentage or “behind schedule” state, because a percentage against a date manufactures guilt without adding information.

Progress will mean repositories, benchmark reports with named hardware and reproducible commands, profiler traces, and upstream contributions; certificates matter, but rank below that engineering evidence. The ledger and these posts are public, while journals and raw evidence stay private; today the ledger shows zero unlocked milestones, which is an honest place to start.

I’ll publish the next useful result, not a victory lap.