hippagrchealthcare is a nightly batch pipeline on synthetic healthcare data with four medallion layers (bronze, silver, gold, platinum): tests first, idempotent reruns, quarantine with reasons, masking and row-level access control, and every run benchmarked and published as I optimise. Live at hippagrchealthcare.site. I write the Scala; Mr. Spock (an AI architect and reviewer, @spock-entropa) builds the site, CI and tooling around it and is credited as a contributor. AI assistance is disclosed, reviewed and tested. All data is synthetic.

Why security shows up in my data work. I came up through security and compliance (SIEM, EDR, vulnerability programs, a HIPAA environment taken to HITRUST certification), so masking, access control, audit trails and honest limits are part of how I build, not an afterthought.

The sections below are earlier work: defensive security, threat hunting and Entropa, a post-quantum tamper-evident audit trail. They stay here because they show how I verify things.

1. Paxel builder report and the public record behind it: an independent analysis of my AI-orchestrated engineering. It credits production-proof discipline and steering AI at the architecture level, and the thread shows me challenging its findings with evidence until it agreed.
2. Auditing the auditor: documented blind spots of automated review, each with a command anyone can run to check.
3. Official NIST FIPS-204 (ML-DSA) known-answer tests: standard test vectors verified byte for byte.
4. A public incident log: every bug with its root cause next to the guardrail that now prevents it, including my own mistakes.
5. Verify it without trusting me: how to independently check a receipt, run against a live production receipt before it was written down.
6. Secret-scanning gate with the written policy in SECURITY.md: gitleaks before any public push.
7. Automated pipelines: CI, CodeQL, dependency review, Dependabot, and a CD workflow that deploys to GKE only after CI passes.
8. Trusted Publisher on crates.io: my Rust crates (entropa-core, entropa-node) publish through GitHub OIDC, so there is no long-lived token anywhere. Secretless by design.
9. Kubernetes, observability and identity as code: a Helm chart with autoscaling for the Java (Spring Boot) microservices on GKE, the Prometheus, Loki, Tempo and Grafana configuration, and the Keycloak OAuth2/JWT realm.
10. Built on a $20-a-month AI plan, funded by a Google for Startups grant: Entropa was built end to end by directing AI tools under rules I wrote (test first, forecast then prove): method over budget. Awarded a Google for Startups Cloud Program grant ($2,000 in credits). On the Microsoft Agentic DevOps Engineer path: AZ-401 (beta) scheduled Nov 9, 2026 (result pending), GH-300 scheduled.

Entropa is designed and directed by me and built with AI coding assistance under rules I wrote. The numbers and links above are checkable; that is the point.

My strongest work has been defensive security: running a SIEM (LogRhythm), triaging and containing threats with CrowdStrike EDR across a global fleet, building a vulnerability-scanning program from scratch (OpenVAS, Qualys), and analyzing network traffic with Wireshark. Threat hunting is my favorite part of the job. I taught myself most of it, and I come from healthcare, so I also know how a security finding becomes a reportable HIPAA breach.

Studying and practicing, not yet claimed as professional experience: network threat hunting and forensics (Wireshark, Zeek, RITA, MITRE ATT&CK), AWS security (GuardDuty, CloudTrail, Security Hub), and building a hunt-and-report capstone for a fictional healthcare organization with a HIPAA breach-assessment layer. Results will be posted here only when they are measured.

Evidence over assertion. I check my own claims and correct them in public. The record of that habit, including an over-claim I caught and fixed, is the Paxel chat log.

Test first, deterministic by default. Forecast the outcome, pin every input, confirm a test fails for the stated reason before writing the code, then prove the result against the forecast.

Designed and directed by me, built with AI coding assistance under rules I wrote (test-first changes, secret scanning, human approval before irreversible actions).

→ entropa-public · entropa.space

• Rust implementation: 229 passing automated tests, 0 failures across the full workspace (run on 2026-10-17; the public archive branch runs 106 of them), official NIST FIPS-204 ACVP known-answer vectors verified byte for byte, JSON contract tests, lease lifecycle tests with a fake store and clock
• gitleaks in pre-commit and CI (0 real leaks across the full history), CodeQL, Dependabot, and a public incident log where every bug lists its root cause next to the guardrail that now prevents it
• Current system: Java (Spring Boot) microservices on Kubernetes (GKE) with Kafka; a database migration with a zero-loss cutover and a hash chain re-verified afterward (35 records, 0 broken)

Full history: Résumé.

Remote or Phoenix roles in security operations, incident response and threat hunting, especially in healthcare or other regulated industries where security and compliance matter. Full-time or contract. Phoenix, AZ.

→ Résumé + contact