Skip to content

ZoaGrad/ColemanWillis

Coleman Willis - Systems Architect

ZoaGrad - Systems Architect

Building deterministic systems for high-stakes environments where "it works most of the time" is not an acceptable operating model.

I work at the intersection of reliability engineering, runtime assurance, AI safety, and distributed systems. The common thread is simple: make fragile systems observable, auditable, and constrained by explicit mechanics instead of vibes.

The Problem

Modern AI and distributed systems are powerful, but they are often opaque and operationally brittle:

  • Observability stacks depend on trained models that can misclassify or hallucinate.
  • Recovery paths are too often "restart and hope."
  • Audit trails are bolted on after the fact.
  • Compliance evidence is difficult to reproduce under stress.

The Approach

I build physics-first and deterministic systems that:

  1. Replace guesswork with mathematics: variance detection, second-order signals, and finite-state correctness.
  2. Make failure visible: forensic recording, transition validation, and replayable incident evidence.
  3. Preserve sovereignty: read-only observation, minimal privileges, and air-gap compatible deployment models.

Active Portfolio

System Role Status
Coherence SRE Variance-based anomaly detection for mission-critical infrastructure v1.0.0, DOI: 10.5281/zenodo.18002927
Air-Node Finite-state agent validation and forensic incident recording Prototype
Blackglass Dojo Adversarial safety testing for AI systems and guardrail failure modes Active
Sovereign Reliability Lab Deterministic phase partitioning under stochastic network conditions Research
Sovereign Embed WASM embedding service for deterministic edge inference Experimental

Verification

The primary systems have been audited for promise-code alignment, edge case behavior, and regression coverage:

  • Coherence SRE: deterministic rolling-window signals, with amplification edge cases verified.
  • Air-Node: forensic trail completeness, with repeated violations preserved and tested.
  • Blackglass Dojo: adversarial safety stack wired end-to-end, with GAMMA scenarios reproducible.

Each hardening pass produced targeted fixes, tests, and merged changes on main.

Why This Matters

High-stakes systems need evidence, not optimism. The work in this portfolio is designed around operational properties that survive contact with real environments:

  • Compliance-ready: NIST-aligned design language, deterministic behavior, and deployability in constrained environments.
  • Auditable: systems produce forensic records and avoid black-box decision paths.
  • Low-risk integration: read-only sidecars and minimal privileged access wherever possible.
  • Practical resilience: signals are designed for operators who need to detect drift before it becomes an outage.

Research & Provenance

Coherence SRE: A Derivative-Based Anomaly Detection Framework for Mission Assurance
Archived at Zenodo, DOI: 10.5281/zenodo.18002927.

Contact

About

Mission assurance, runtime assurance, and reliability systems portfolio for Blackglass Continuum.

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors