ExNulla Site: static-first proof-of-work portfolio
Astro + Nginx portfolio system with atomic releases, isolated demo artifacts, provenance-aware builds, and operational guardrails designed to survive real updates.
astronginxatomic-releasescicd
James Huson · Applied AI Systems Engineer
I build the systems around the model.
I design and ship applied AI systems that connect models to real workflows while preserving deterministic validation, security boundaries, auditability, and operational control. My work spans agent orchestration, governed tool execution, developer infrastructure, and production delivery.
Positioning
Applied AI platform and systems engineering across TypeScript, Node.js, Python, PostgreSQL (schema design, production operation), the Anthropic Claude API, CI/CD, cloud deployment, schema-gated agent workflows, policy/authorization engines, threat modeling, and operational verification. 20+ years of technical/operations experience overall; hands-on software and applied AI systems engineering since early 2026. I am strongest where architecture, implementation, and commercial constraints have to agree.
What I do
- Applied AI systems: design agent workflows, tool boundaries, context flows, validation paths, deterministic fallbacks, and audit surfaces.
- Platform delivery: TypeScript strict, Node, Python, PostgreSQL, Astro, Nginx, GitHub, DigitalOcean, Cloudflare, CI/CD, and boring deploy discipline.
- Architecture judgment: frame the problem, lock constraints, define boundaries, choose the smallest useful build path, and make trade-offs explicit.
- Engineering governance: schema gates, prompt/version discipline, threat tests, provenance, human review, and reproducible delivery around model output.
TypeScriptNode NginxAstro DigitalOceanCloudflare PythonPostgreSQL Anthropic Claude APIMCP Policy enginesIAM / RBAC Agent systemsAI governance
Operating principles
- Start with the problem, the constraints, and the economics—not the tooling.
- Prefer static-first, composable systems over heavy stacks with unclear operating cost.
- Make deployment boring, rollback cheap, and provenance visible.
- Prototype fast, then harden deliberately when the design proves it deserves to live.
- Use AI aggressively, but keep authority, judgment, and accountability human.
ExNulla is the reference implementation of how I work: deliberate scope, visible proof, repeatable delivery, and a bias toward systems that survive contact with reality.
Background
- Operator background: I spent 20+ years designing and building businesses, operations, and revenue systems before formalizing that experience in software and AI systems.
- Working style: I define the problem, design the system, build the assets, test the weak points, and improve what breaks.
- ExNulla: a public engineering portfolio and live reference system spanning architecture, repositories, CI/CD, infrastructure, demos, and documentation.
- Proof: the work is documented, shipped, inspectable, and increasingly open-source.
The result is a technical portfolio that can be inspected instead of taken on faith.
Range
- Software / infra: TypeScript strict, JavaScript, Node.js, Astro, Vite, Nginx, Git/GitHub, PostgreSQL (schema design, production operation at scale), SQLite, GraphQL exposure, Rust in active learning, Docker-aware CI/provenance thinking.
- Cloud / delivery: DigitalOcean, Cloudflare, Vercel, static deploy architecture, artifact isolation, version stamping, CI/CD hygiene.
- Creative / tooling: Blender, SketchUp, 3D CAD, Premiere Pro, DaVinci Resolve, documentation systems, spreadsheet and financial workflow tooling.
- AI stack: Anthropic Claude API, LLM application architecture, MCP (Model Context Protocol) integration, agent workflows, policy/authorization engines (identity, RBAC-style approval and delegation, replay-attack resistance), prompt/version control, schema-gated outputs, tool integration, evaluation, and deterministic multi-agent build workflows.
What I am optimizing for
- Target roles: Applied AI Engineer, AI Platform Engineer, AI Solutions Architect, Agent Infrastructure Engineer, Forward-Deployed AI Engineer, or Staff Software Engineer.
- Value to a team: bridge product, operations, business constraints, security, and implementation without losing the thread between architecture and delivery.
- My edge: I think in systems that must be built, operated, explained, and made economically viable.
I am strongest where architecture, execution, and commercial reality all have to agree.
Case studies (curated)
Stream Ops: deterministic live production stack
OBS, Multi-RTMP, Streamer.bot, browser overlays, generated media, and operator runbooks turned into a repeatable production system instead of a pile of fragile settings.
obsautomationopsrunbooks
Space Server: 50 kW orbital compute trade study
Constraints-first architecture work on orbital AI infrastructure: power budget, GPU packing, radiator physics, launch packaging, and the discipline to separate speculation from locked assumptions.
thermalsystemsarchitecturecompute
Nexus: policy-aware action router + authority governance layer
A private-runtime authorization gateway for AI agents: identity and vault-backed credential handling, a governed lexicon/compile layer, an MCP adapter, and a 10-scenario policy test suite covering approval workflows, delegation limits, replay-attack detection, and default-deny access control.
policy-engineiamrbacmcpagent-governance