LW
pdf

Sydney · Australia  ·  for Neilson Financial Services

Liam Wynne

Data & AI Builder

I don’t present ideas — I build them. Civil engineer turned AI engineer. I ship LLMs into production, wire systems together with webhooks and APIs, and use Claude Code in my actual workflow every day — not as a party trick.

Portfolio — for this role

ERD Studio in action — designing a data warehouse on a canvas inside VS Code

ERD Studio

vs code extension · open source · on the marketplace

Visual ERD designer for dbt. Design your data warehouse on a canvas, in your repo, where Claude and other agents can read and edit it via Model Context Protocol. TypeScript.

for Neilson — if dbt is in (or near) your stack, this is the same accelerator that took my team from “we need a new model” to production in days, not weeks. The one I’m most proud of.

Live demo — Anthropic Managed Agents cloud session running the sales-quoting orchestrator, showing the agent loop, six specialists, and the raw event stream

Sales Quoting Agent

multi-agent claude poc · cloudflare worker · anthropic managed agents

Orchestrator + six specialists (RFQ parser, rate lookup, surcharge engine, client terms, margin checker, quote drafter). Local MCP server, Cloudflare Worker, and the same agent loop deployed on Anthropic’s Managed Agents cloud runtime. Rate limits, cost ceilings, 15 passing tests.

for Neilson — directly the shape of the JD: “agents, pipelines, automation” against real business logic, with three runtimes proving the same prompts work laptop → edge → cloud.

The Big Red roster timeline — coloured time-slot bars across the day showing when each staff member is available between their fixtures

Big Red

next.js on cloudflare workers · live · api integration

A small café roster planner. Pulls live football fixtures from the SSFA API and shows when each casual staff member is actually available on a Saturday. Next.js + Cloudflare Workers (via OpenNext) + KV. Shipped because the cross-referencing was annoying me.

for Neilson — the JD’s “ship a lightweight internal tool without waiting for a designer”. Frontend + API integration + deploy, end-to-end, in a weekend.

The liam plugin marketplace grid — 12 plugins for Databricks, Power BI, Atlassian, SharePoint, messaging, SSH, status line, and more

liam

claude code plugin marketplace · 12 plugins

Personal Claude Code marketplace covering Databricks admin, Power BI local editing, Atlassian + SharePoint REST, multi-platform messaging, Raspberry Pi SSH, status line. One-line install: /plugin marketplace add liam-machine/liam.

for Neilson — the JD’s “AI-native tooling as a daily multiplier, not a party trick” line. Tooling I built for myself and packaged so other engineers can install it in one command.

Work

AI / Data Engineer

John Holland Group

2022—present

  • Built and shipped ERD Studio — an open-source VS Code extension on the marketplace that uses LLMs to compress dbt data-warehouse delivery from five months for a whole team down to two weeks for one engineer with AI.
  • Shipped a Retrieval-Augmented Generation pipeline on Delta Live Tables — vector search over unstructured operational content, wired into an enterprise chatbot.
  • Automated Power BI measure generation with LLMs — turned days of manual DAX work into an integration that writes and validates measures in minutes.
  • Use Claude Code as a daily multiplier — agentic coding, custom skills, MCP integrations. Built personal tooling around it, not just consumed it.
  • Wired systems together end-to-end — GitHub Actions, dbt Cloud, Databricks workflow APIs, webhook triggers and event-driven jobs. Manual processes become hands-off pipelines.
  • Designed full-stack data workflows across AWS, dbt, Databricks and Power BI — ingestion, transformation, modelling, reporting.
  • Modelled data with bronze / silver / gold architecture — aggregated, scalable, observable.
  • Built high-performance DirectQuery KPI dashboards in Power BI — raw operational data to executive summary with minimal latency.
  • Deployed MLflow pipelines for classification — continuous training, registry and governance.
  • Engineered ML pipelines for Tunnel Boring Machine telemetry on AWS Glue, SageMaker and Airflow — real-time sensor streams to insight.
  • Designed and implemented the team’s dbt CI/CD — GitHub Actions pipelines on every PR, environment promotion, automated tests, deploys into Databricks. Live in production.
  • Presented an AI primer to ~80 IT staff and ran brownbags on AI-assisted data engineering for the wider team.

Systems Engineer

John Holland Group — Sydney Gateway

2018—2022

  • Systems Lead on a multi-year CBD infrastructure project.
  • Built and rolled out Microsoft Power Platform tooling (Power Apps, Power BI, Power Automate, SharePoint) across the site — workflow automation before it had a marketing department.
  • Digitised manual workflows end-to-end and trained the team into the new systems.
  • Administered the project quality software (PPW) and produced the analysis that drove decisions.

Site Engineer

CPB Contractors — WestConnex New M5

2015—2018

  • Site Engineer on Stage 2 of the New M5 motorway tunnel.
  • Ran site setup, decline excavation and structural steel install at Arncliffe.
  • Owned daily reporting, program updates, cost forecasting and material procurement.
  • Interfaced with designers and supervised the field workforce — safety-critical operations, no room for sloppy data.

Stack

Claude Code
agentic dev, daily
Python, SQL
daily
TypeScript
extensions, tooling
LLM APIs & MCP
anthropic, openai, model context protocol
Databricks
lakehouse, delta live tables
dbt
transformation, modelling
AWS
glue, sagemaker, airflow
GitHub Actions
ci/cd, webhook automation
MLflow
model ops
Power BI / Power Platform
analytics, automation

Education

2018—2020

Master of Computer Science

University of Wollongong — AI & Big Data

Distinction

2011—2014

Bachelor of Civil Engineering

University of Wollongong

First-class honours

Contact

Email
[email protected]
Phone
•••• ••• •••
Location
Sydney, Australia
GitHub
liam-machine
References on request