Distributed systems, AI governance, security research, and full-stack platforms. All deployed on Cloudflare.
Red-team security research demonstrating AI can fabricate a believable organization with 9 synthetic executives for under $250. Includes an 8-signature detection engine with per-signature confidence scoring and a full defense playbook.
Research basis: 3 publications • Validates propositions 12–16 • Powers all 4 solution modules
A 4,613-line TypeScript orchestration engine coordinating 6+ concurrent AI agents through a four-stage helical pipeline (FORGE, DEFINE, BUILD, VERIFY) with DAG-based scheduling, tier-priority queuing, and gradient auto-compaction.
Research basis: PUB_01 • Orchestration layer for cross-model verification (proposition 31)
Seven-pillar AI governance framework with specialized Durable Object agents evaluating proposals through structured debate-vote-implement cycles. Deployed on Cloudflare Workers with 7 DO classes, 6 KV namespaces, and Workers AI binding.
Implements disagreement visibility (proposition 33) • Governance layer for Helix Fabric decision cycles
Model Context Protocol servers exposing calibrate, evaluate, and attend tools enabling any LLM to operate within constitutional guardrails. Drift scoring 0-100 with crisis detection routing. Core principle: refuse rather than hallucinate.
Implements measurable credibility (proposition 34) • Constitutional layer for Edenic Legion & Portal Chat
HIPAA-scope B2B healthcare chaplaincy platform on Cloudflare Workers via OpenNext. Part of a multi-tenant unified Worker serving 5 product domains with shared D1, Vectorize, and Workers AI bindings.
Trust-based domain application • Demonstrates sector exposure (proposition 18) in healthcare
Four-source parallel RAG pipeline across 5 production Vectorize indexes with priority-based context merging, crisis content escalation, and 8,000-token budget management with provider-agnostic fallback across Workers AI, Groq, and Gemini.
Semantic layer for Helix Fabric Overseer • Enables multi-signal detection (proposition 23) via Vectorize embedding
Six-agent governance chat (Proposer, Critic, Auditor, Redteam, Sanitizer, Firewall) with fail-closed semantics, NIST CSF 2.0 compliance across 106 subcategories, and hash-chained audit logs with BM25 search.
Implements signed outputs & audit trails (proposition 32) • Governance interface for Mantle constitutional layer • Audit architecture for provenance infrastructure