Greater Lincoln, UK · Remote

Matt “Foxx”
Treacy

Solution & Software Architect / AI-Native Engineer

Systems architect and AI-native engineer with ~20 years across IT operations, DevOps and cloud — building hands-on with LLMs since GPT-3 early access, four years before most teams started.

~20
Years · ops → AI
4+ yrs AI
LLMs since GPT-3
~1 mo
Core to prod, solo
Remote
Lincoln, UK · remote
01About

The short version

AI expertise, built on two decades of operations judgement.

The rare combination: someone who knows what breaks in production and has been building with LLMs since before it was a category.

I'm a systems architect and AI-native engineer with nearly two decades across IT operations, DevOps and cloud — and I've been building hands-on with large language models since GPT-3 early access, four years before most teams started.

I architect like someone who's been paged at 3am: I care about isolation, scaling, cost and what actually breaks in production — not just what demos well. Recently I designed and built, solo, a secure multi-tenant platform end to end: schema-per-tenant data isolation, RBAC/ABAC access control, an AI app-generation system (plain-English brief → working app), safe sandboxing of untrusted code, and a single AI gateway for model governance. I took the core to production in about a month and evolved it from there.

The difference isn't the tools — it's the judgement underneath them. Twenty years of operations is why the systems I build hold up: I know how compute and storage scale under load, how multi-tenant isolation fails, and what an architecture really costs over three years rather than three weeks.

AI-native development is also leverage. I run several workstreams in parallel — the output of a small team, with one owner's architectural judgement on every branch. And I'm deliberately not wedded to any stack, language, product or tool: I pick what's right for the problem and keep hunting the next, better way to build. The tools are interchangeable; the judgement is the constant.

I'm always in a state of discovery, and right now I'm most energised by the intersection of deep cloud/ops experience and AI-native software design — and by helping teams and whole organisations adopt the way of working I've spent four years refining.

I architect like an engineer who's been paged at 3am — not someone who just picked up the tools.

Matt "Foxx" TreacyArchitect & AI-native engineer

02Experience

Where I've worked

2018 — Present

Current

Senior IT Systems Engineer → Architect-level work

Adaptavist

Operating well beyond the title: architecting and shipping production systems solo, from multi-tenant AI platforms to self-service Kubernetes infrastructure.

  • Sole architect & engineer of a secure, serverless multi-tenant AI platform — core to production in ~1 month, then evolved solo from there.
  • Designed schema-per-tenant data isolation, RBAC/ABAC access control and a single AI gateway for model governance, failover and cost control.
  • Built an agentic AI app-generation system: a plain-English brief becomes a validated, deployable multi-file app via an LLM tool-calling build loop.
  • Core engineer on a self-service container platform on AWS-managed Kubernetes (EKS) — autoscaling compute and storage, multi-tenant isolation, full image-to-UX ownership.

2007 — 2018 · 11 years

Founder & Director

NXS Computers

Founded and ran a computer-services business end to end — left having outgrown the client base, wanting enterprise scale and to push the industry forward.

  • Owned everything: delivery, client relationships, operations and the bottom line.
  • A decade of hands-on operations across hardware, networks and systems — the production instinct that underpins how I architect today.
03Case studies

Selected work

Case 01

Enterprise Multi-Tenant AI Platform

Sole Architect & Engineer

Built to turn non-technical, no-code users into safe, self-serve app builders: they describe what they want in plain English and get a working, deployable, database-backed web app — inside guardrails that keep it secure and compliant by design. A serverless multi-tenant platform; core shipped to production in ~1 month, then evolved solo.

  • The mission: democratise software — empower non-technical and no-code users to build and deploy their own apps in a guardrailed, secure-by-default environment, no engineer required.
  • Multi-tenancy: schema-per-tenant data isolation with per-tenant object-storage partitioning — tenants isolated by design.
  • AI app generation: an agentic build loop (LLM tool-calling that writes, patches and validates files over a multi-step budget) turning a natural-language brief into a deployable app.
  • Safe extensibility: a runtime for running untrusted, AI-generated code in isolation, exposing a governed SDK instead of raw platform access.
  • Enterprise-ready: SSO, end-to-end audit logging and policy-driven RBAC/ABAC — plus a credentials pattern where apps hold only an opaque handle (secrets resolve server-side, outbound calls via a proxy).
  • Governance & scale: all model access routed through a single AI gateway for failover and cost control; fully serverless from prototype to enterprise.
Next.jsTypeScriptPostgreSQLServerlessLLM agentsAWS

Internal, unreleased product — described at patterns altitude, no product name, no screenshots.

Case 02

Self-Service Kubernetes Container Platform

Core Engineer — AWS / EKS

A self-service platform letting consultants provision and orchestrate their own containers on AWS-managed Kubernetes (EKS), fronted by Rancher. Hands-on at every layer.

  • Cluster & workloads: EKS running multi-tenant container workloads, with Rancher as the management/UI layer for self-service orchestration.
  • Compute scaling: autoscaling cluster compute so workloads expand and contract with demand.
  • Storage scaling: persistent, scalable storage provisioned to match tenant usage.
  • Full-stack ownership: image build/layering and container packaging through to deployment config and the consultant-facing experience.
Kubernetes (EKS)AWSDockerRancherDevOps
04AI-native methodology

How I build

A four-year-refined practice, not a recent pivot. How I build production systems with AI in the loop — kept at the level of patterns and judgement, never specific to any one product.

01

Operations judgement first

Every design starts from how it fails: isolation boundaries, failure modes under real load, and three-year cost — not three-week demos. Twenty years of being on-call is the durable, non-commoditising part.

02

AI as a force multiplier, governed

Model access routed through a single gateway for governance, failover and cost control. AI accelerates the work; the architecture decisions stay deliberate and owned, not delegated.

03

Isolation by design

Multi-tenancy and untrusted-code execution treated as first-class: schema-per-tenant data isolation, sandboxed runtimes, and governed SDKs instead of raw platform access.

04

Agentic build loops

LLM tool-calling that writes, patches and validates over a bounded multi-step budget — turning a plain-English brief into a deployable, validated artefact, with humans owning the architecture.

05

Serverless for velocity, AWS where it fits

Serverless for DX and speed; deliberate use of AWS where it's the better tool. A considered trade-off backed by deep cloud/DevOps experience — not a default.

06

Ship the core, then evolve

Get a real, secure core to production fast, then evolve it solo. Working software beats slideware — and it's the strongest proof you can hand a hiring team.

07

Pragmatic, not dogmatic

Not wedded to any stack, language, product or tool — I pick what's right for the problem and keep hunting the next, better way to build. AI fluency makes language a non-constraint, so the choice is always judgement, never habit.

08

Parallel, agentic throughput

Several workstreams running at once, with architectural judgement held on every branch. The leverage of AI isn't speed for its own sake — it's the output of a small team with one owner accountable for every decision.

09

Methodology that travels

The way of working is an asset, not a personal trick. I install it in teams and whole organisations — bringing people up the AI-native curve so the leverage compounds beyond any one engineer.

05Skills

Toolkit

Architecture

  • Solution Architecture
  • Software Architecture
  • System Design
  • Multi-tenant SaaS

AI / LLMs

  • LLMs / Generative AI
  • AI Agents
  • Agentic Systems
  • Model Governance

Cloud & Platform

  • AWS
  • Kubernetes (EKS)
  • Serverless
  • Docker
  • DevOps
  • Infrastructure

Build stack

  • Next.js
  • React
  • TypeScript
  • Node.js
  • PostgreSQL
06Engagements

How I can help

Whether you need someone to own the architecture end to end or to level up how your team builds — here's where I add the most leverage.

Full-time

Founding / Principal Engineer

Own the architecture and ship the hard core. I take an ambiguous problem to a secure, production-grade system — and stay accountable for every decision along the way.

  • 0→1 platform & product architecture
  • Hands-on, solo or leading a small team
  • Production-grade from day one — isolation, scale, cost
Contract

Platform & Solutions Architecture

Design and de-risk the system before it's expensive to change. Multi-tenant isolation, cloud cost, AI governance — the decisions that are painful to reverse.

  • Architecture & design reviews
  • Multi-tenant, cloud & AI-gateway design
  • De-risking the decisions you can't easily undo
Fractional / Advisory

AI Enablement & DevEx

Install the AI-native way of working across your teams. Four years of refined practice, transferred — so the leverage compounds beyond any one engineer.

  • Agentic build loops & parallel workflows
  • Guardrails, governance & model strategy
  • Coaching teams up the AI-native curve
07Contact

Let's talk about the right role.

Founding engineer, platform / solutions architect, or AI-enablement & DevEx leadership — installing the AI-native way of working across teams and orgs. If it pays for judgement, I'm interested.

Full-time · Contract · Fractional / Advisory

Prefer email?

Hidden from scrapers — click to show.

Elsewhere