Clockwork — AI Integration Architecture

Sound familiar?

The AI gap.

The question isn't whether to use AI. It's how — and nothing you've tried so far has stuck.

AI experiments stuck in silos. Proof-of-concepts that never reach production. Individual teams trying things, nothing connecting.

Unpredictable costs. API bills that spike without warning. GPU spend that's hard to forecast. No clear picture of what AI actually costs to run.

Vendor lock-in anxiety. Tied to one provider's API, one model, one platform. No exit strategy. No flexibility when something better arrives.

Security and governance gaps. Data flowing to third-party APIs with unclear policies. No framework for what's safe to automate and what isn't.

Neural network of golden interconnected nodes

Approach

Not another AI consultant.

I build working systems, not strategy documents. The difference matters when the hype fades and you need something that runs.

AI as infrastructure

Models aren't magic — they're components. Like databases, queues, and APIs, they belong in your architecture. I help you put them there properly.

Capability, not dependency

Every engagement is structured to transfer knowledge, not create reliance. You own the capability when we're done. Your team runs it, not mine.

Built with what I sell

I use AI to build AI solutions. This isn't theoretical knowledge — it's daily practice. The tools I recommend are the tools I use.

Services

Three layers. One path forward.

You can start wherever you are. Each stage builds naturally into the next.

Understand

AI Readiness & Education

Cut through the noise. Understand what AI actually is, what's real, what's coming, and what it means for your business specifically.

Executive and team AI literacy workshops
What LLMs can and cannot do — honestly
Opportunity identification and risk evaluation
Where automation breaks and where it works

Strategise

Integration Architecture

Map where AI fits in your systems. Not a generic roadmap — a specific, technical plan built around your stack, your data, and your goals.

System and workflow analysis
Model selection and architecture design
Cloud, local, and hybrid strategies
Cost modelling and build-vs-buy analysis

Build

Implementation & Transfer

Hands-on building, alongside your team. Working prototypes, production systems, and the knowledge to maintain them without me.

Prototype and proof of concept development
Production AI pipeline engineering
API integration and orchestration
Team training and knowledge transfer

Brass clockwork gears merging with circuit board traces

Proof of work

Built. Shipped. Running.

A selection of projects that demonstrate real AI integration architecture — from vision pipelines to multi-agent systems to GPU orchestration.

Vision AI + Cloud Infrastructure

TagMills

Built TagMills, a scalable batch image-processing platform that uses vision-language models and workflow orchestration to generate structured metadata at volume. It supports automated captioning, tagging, classification, and content enrichment with modular processing stages designed for throughput and downstream asset pipeline integration.

FastAPI · RunPod · Vision-Language Models · Scheduler · React

Autonomous AI Agent

Justabot

Created Justabot, a modular AI agent framework that connects LLMs to tools, APIs, automation routines, memory systems, and device control interfaces. It is an execution-oriented agent layer for bounded real-world tasks with persistent context, auditable actions, and extensible tool integration.

Claude · MCP · Node.js · Tuya IoT · Ollama · ComfyUI

GPU Image Generation Platform

Meld

Built Meld, a multi-model orchestration layer for AI workflows that coordinates prompt routing, tool invocation, workflow state, and generation pipelines across local and remote models. It serves as a control plane for complex creative and automation tasks, integrating ComfyUI-style workflows, external APIs, and structured task execution.

ComfyUI · Docker · RunPod · FastAPI · React · Konva

Multi-Agent Architecture

Clockwork Cognitive Council

Multiple AI agents with distinct cognitive profiles debate and synthesise insights through a modular architecture. Used for scenario modelling, decision analysis, and adversarial testing of ideas. Connected via ZMQ message passing.

Python · ZMQ · Multi-LLM · Graph Activation · Modular Architecture

Creative AI Pipeline

Muage

Developed Muage, an automated media pipeline that transforms timestamped lyrics and music structure into synchronised image and video sequences. It combines text parsing, scene generation, prompt construction, image synthesis, refinement stages, and video assembly into a reproducible end-to-end workflow for music-driven visual production.

Python · Flux.1 · ComfyUI · FFmpeg · Lyrics Processing

AI Memory Systems

Flux Memory

File-based external memory system for AI agents. Persistent, structured, and semantically aware — giving AI systems the ability to remember, reflect, and build on previous interactions across sessions.

JSON · Semantic Indexing · File-based Architecture · LLM Integration

Background

Infrastructure roots. AI present.

Thirty-two years building the systems businesses depend on. Started on SCO UNIX in 1993. Solaris, HPUX, AIX, Linux — if it ran a shell, I've administered it. From rack-and-stack to platform architecture, from DR implementations to virtualisation strategy.

Enterprise infrastructure across banking, telecommunications, healthcare, government, higher education, and aviation. Mission-critical systems, every time. The kind of environments where downtime isn't theoretical.

Now two years deep in AI — not just using tools, but understanding how models work, how to orchestrate them, how to build production systems around them. Forty-plus projects and counting. Building daily.

Few people have spent decades in infrastructure and then gone deep on AI systems engineering. I speak both languages — because AI is the new middleware, and models are the new microservices. The same engineering discipline that made infrastructure reliable is exactly what AI integration needs.

Finance Telecommunications Healthcare Government Education Aviation eCommerce Blockchain

Years in infrastructure

40+

AI projects built

Production AI systems

Slide decks sold