AI Platform Engineer (release gating, self-hosted inference)
Open to mid-level and senior IC roles, W2, remote (US).
I build release gates that block an AI deploy when its safety checks fail.
deploycontrol PASSplanted fault FAILgatereceipt
Requirements and proof
| Requirement | Proof |
|---|---|
| Ship changes through a deploy pipeline with checks that block a bad release | Release-gate case study |
| Plan a reversible deployment, with a rollback path, for a critical system | COBOL-to-Rust migration case study |
| Investigate an incident to root cause and ship a durable remediation, not a workaround | Deterministic gate battery case study |
| Define reliability signals that block a regression before it reaches customers | Release-gate case study |
| Reduce single points of failure through automation rather than manual operational work | Agentic fleet verification case study |
| Keep a reproducible, auditable record of what was deployed and verified | Artifact-lineage case study |
What could go wrong hiring me
Most proof here is self-built, not incident response on a live, always-on system.
Counter: COBOL-to-Rust migration case study
On-call and incident-command history here is not documented with employer pager records.
Reliability and gate figures are self-reported, not published by a third-party status-page provider.
Counter: Public adversarial-gate repo with CI
Verify in 60 seconds
- Clone the public COBOL probe, run python probe.py, and check that its synthetic loss figure matches the README, https://github.com/mpuodziukas-labs/cobol-pic-probe
- Clone the public adversarial-gate repo and check the CI-green test count against its README, https://github.com/mpuodziukas-labs/llm-adversarial-gate