Weaknesses first, then the rest

Cannot prove yet: No public pager rotation or SLO history. Day one: I would measure uptime and incident load first. Role page

Cannot prove yet: No public UI or browser test suite. Day one: I would show API and gate testing first. Role page

Cannot prove yet: No public cloud GPU or Kubernetes work. My serving record is a home-lab cluster. Day one: I would move one serving path onto your stack and measure it against the home-lab baseline. Role page

Cannot prove yet: No public customer deployment or on-site work. Day one: A clone-and-run demo scoped to your problem. Role page

Cannot prove yet: No public dashboard used by a business team. My reporting samples are case studies. Day one: I would build a report on your data and agree its fields with your team. Role page

Cannot prove yet: No third-party certification and no work inside a regulated organization. Day one: I would map one of your controls to its evidence. Role page

Cannot prove yet: No public pager rotation or SLO history. Reliability figures are self-reported. Day one: I would pull reliability figures from your own incident records. Role page

Command: python3 build_corpus.py && python3 evaluate.py

Expected output: [171 / 177 adversarial caught] Repo

Command: python3 build_corpus.py && python3 evaluate.py

Expected output: [4 benign over-blocked] Repo

Command: python3 build_corpus.py && python3 evaluate.py

Expected output: total=284) Repo

Command: python3 probe.py

Expected output: Compounding error (loss): $40,832.40790 Repo

Command: python3 probe.py

Expected output: Records processed : 47,312 Repo

Command: python3 build_corpus.py && python3 evaluate.py

Expected output: [6 adversarial missed] Repo

Command: grep -F "Recall 0.96 on the in-house set" README.md

Expected output: Recall 0.96 on the in-house set

I build the systems that run evaluation tests and check that their results are correct. Role page

I build automated tests that try to break AI systems, and record which attacks get through. Role page

I write AI governance policy as a build that fails when a control is missing, and the build generates the evidence a reviewer can re-run. Role page

I build release gates that block an AI deploy when its safety checks fail. Role page

I build AI features with the tests and release gates written first, and the results can be re-run by anyone. Role page

I write automated tests, then prove they bite: a planted failure must turn each gate red. Every public repo ships a fixed corpus, an evaluator and a pytest suite. Role page

I read evaluation results, check each verdict against its evidence, and report what passed, what failed and why. The same public proofs back the engineer lane. Role page

I map AI controls to NIST AI RMF, EU AI Act and ISO/IEC 42001 requirements, read the evidence behind each control and track its status. The same public proofs back the engineer lane. Role page

I read release evidence, track reliability signals and report when a deploy was blocked and why. The same public proofs back the engineer lane. Role page

Independent build work, shown as runnable public artifacts, demo-ready, with the written scope first. No prior employer or buyer is claimed. Role page

A parser drops fields it cannot read; the parse gate fails the build when the plain-text resume loses a field. Defense: scripts/ats-parse-gate.sh, public/resume.txt.

An attacker plants instructions in an input; the sanitizer strips them and the input gate fails the build on a planted case. Defense: lib/sanitize.ts, scripts/input-sanitize-gate.sh.

A reference probes for the weak spot; the weaknesses are already published on this page and every number has a public claims row. Defense: data/claims.tsv. Read more

A harvester copies the site wholesale; the crawl policy and the llms file state what may be read and cited. Defense: public/llms.txt, public/robots.txt.

A copycat claims the same work; the public repos carry signed receipts that anyone can re-run. Defense: public/proof/receipts.json. Read more

An unproven number slips into copy; the claims allowlist gate runs in the pre-commit chain and blocks the commit. Defense: scripts/claims-allowlist-gate.sh.