FAQ
Frequently Asked Questions
Every claim on this site is anchored in the proof record.
Why This COBOL-Forensics × LLM-Adversarial × PQC Overlap Is Rare
Replay-verified outcomes + adversarial validation across COBOL, AI, and post-quantum domains
One operator who pairs replay-verified COBOL forensics with a forensic-grade audit trail for model-risk review and a replay command per finding.
Every answer below closes a decision gap. Addressing the gap early means the fix and a replayable test are already in the artifact trail — not a guarantee of any examination outcome.
Common Questions
Are you available?
Open to one Staff/Principal role, Q3 2026. The fastest way in is the diagnostic: three questions that surface whether a gap is active before any call.
How do you work with a team?
Written scope first, milestone-gated, replay command on every finding. The first artifact lands inside 48 hours. You reproduce each number from the artifact it ships with.
Do you sign an NDA?
I work under NDA on your infrastructure — code and data stay with you; you get artifacts, replay commands, and a signed audit log.
Do you work remote?
Remote-first. Inference runs on a sovereign on-device cluster: zero managed-API calls, zero data egress. Nothing in the pipeline leaves a controlled environment. Source, build, and secrets stay on your infrastructure; the only thing that crosses the wire is a signed artifact with its replay command.
When can you start?
I start fast: no onboarding queue, no ramp period, no first-week-of-meetings. The first artifact lands in days, not sprints.
What does this actually save my company?
A defect three years of clean audits missed, found in 127 minutes and reproducible by anyone. A $26,400/mo inference bill cut to $0 on hardware I own. Every number here is something you or your team can re-run, not something you have to trust.
Documented Failure Modes
- COBOL PICTURE-clause truncation — forensic audit methodology, replay-verified
- AI evaluation gaps — adversarial validation, deterministic audit trails, forensic-grade documentation
- Post-quantum exposure — CNSA 2.0, FIPS 203/204/205, 2030 migration deadline (NIST IR 8547)
- LLM adversarial injection — 0 false negatives across the public adversarial corpus (140/140 caught, n=220)
- FFIEC examination findings — arithmetic boundary in payment pipeline
- SR 11-7 model validation — COBOL upstream audit trail for regulated models