About
I am a platform architect specializing in building and operating production LLM-agent infrastructure, currently working at Visionary Solutions since 2018 with progressive roles culminating in Platform Architect. With over six years of experience, I have designed multi-layer agentic engineering platforms and led IR&D on zero-knowledge proof systems for privacy-preserving verification.
My work spans TypeScript, Python, Rust, and Go, building multi-agent Claude Code pipelines, custom MCP servers, and trace-first observability products published to npm. I develop secure configuration tooling, Kubernetes and Terraform infrastructure, and formal verification harnesses including Lean 4 machine-checked proofs and property-based simulation, applied in high-security federal environments and production AI-agent workflows.
Outside work: distance running with a DC run club, and gaming.
What I'm looking for
Senior platform and AI-agent infrastructure work, fully remote (US, based in Washington, DC). Part-time, contract, freelance, or fractional engagements: W-2 or 1099 through my LLC. Strongest fit: agent tooling and developer infrastructure, platform engineering with formal verification wired into CI, and privacy/ZK systems. I want teams that value evidence over assertion: proofs, traces, and evals in the loop.
Experience
Platform Architect
Visionary Solutions
Jan 2021 – Present
- Architected AI-augmented engineering workflows in production: multi-agent Claude Code pipelines with custom MCP servers (IPC, token proxy) and a cross-model adversarial review gate (Claude/Codex/Gemini) that blocks completion until independent-review findings are addressed.
- Led IR&D on CHAP, a STARK-based zero-knowledge proof system for privacy-preserving financial eligibility verification.
- Platform architecture for unclassified federal programs; Zero Trust and multi-level security patterns.
Claude CodeMCP serverZero TrustConfiguration managementInfrastructure automation
Systems Engineer
MTSI (Modern Technology Solutions, Inc.)
Jan 2018 – Jan 2021
- Systems integration and configuration management for classified (Protection Level 3) federal programs: automated configuration management and implemented access controls across integrated systems.
- Program support and R&D across the system lifecycle.
Systems integrationConfiguration managementAccess controls
0rrery: trace-first observability platform for AI-agent workflows (Claude Code, Codex CLI): HTTP trace ingest, SQLite storage, live WebSocket view, cost and failure analytics, wired into nine agent lifecycle hook events.
Single-process local-first design. Published to npm with provenance.
rtk: Rust CLI proxy for agent workflows: rewrites and filters tool commands behind a deterministic allow/deny/ask protocol, cutting agent token spend 60–90% while enforcing policy before execution.
@0pon/correctness: six-tier verification harness (tsc → Zod → Quint model checking → Lean/Dafny proof → property-based simulation → DeepEval LLM evals) behind one CLI, run in cost order; distributed via a private Claude Code plugin marketplace and used across a real product portfolio.
n0de: attribute-based access control for compartmented workloads: non-interference machine-proven in Lean 4, enforced by a Rust/Cedar policy engine with PostgreSQL row-level security and SELinux, every layer differentially tested against the proof (17 CI gates). Core of a Navy SBIR Phase 1 submission. Directly applicable to multi-tenant agent isolation: access decisions that are provably, not plausibly, correct.
Production client platforms: delivered solo under 0pon: property-management/trust-accounting platform (double-entry ledger, ACH), DC transit data platform (TimescaleDB/PostGIS), HIPAA-posture referral SaaS with Claude-vision document intake, Supabase-RLS CRM product line.
RustSQLitePostgreSQL
Education
Western Governors University
M.S. · Cybersecurity
2022
Western Governors University
B.S. · IT Management
2020
Skills
GoRustPythonTypeScriptABACZero TrustCloudflare WorkersPostGISTimescaleDBPostgreSQLCI/CD designGitOpsTerraformKubernetesToken optimizationDeepEvalAgent ObservabilityClaude CodeModel Context ProtocolLLM orchestration
Languages
English (Native or bilingual proficiency)
