AI Engineer in Sydney, building the agentic networked economy.
I'm an AI Engineer at Decidr.ai, where we build the infrastructure behind the agentic networked economy: agentic AI systems that don't just automate tasks, but reason, adapt, and act in real time.
What I bring is the full stack of production AI: not just the LLM pipelines, retrieval, and agent orchestration, but the infrastructure that keeps them running. Before Decidr I worked as a DevOps Engineer at Nexl, and spent years as a software engineer shipping products end-to-end. That range means I can take an AI idea from prototype to a monitored, scalable service.
I hold a Master of Data Science & Innovation from the University of Technology Sydney, and I write regularly on machine learning and applied AI.
Education
M.Sc. Data Science & Innovation
University of Technology Sydney · 2024
Certifications
- Deep Learning Specialization
- Sequence Models
- Structuring ML Projects
- Apache Airflow
Selected Work
Open-source tools I've built to solve real problems: a sample of the engineering range I bring to production AI systems.
LCP Debugger
Chrome Extension · JavaScript · Web Vitals API
A Chrome extension for debugging Largest Contentful Paint. Identify performance bottlenecks in real-time, right in the browser.
View Source →Pt command line
Ruby · CLI · Pivotal Tracker API
A command-line interface for Pivotal Tracker. Manage stories, sprints, and velocity without leaving the terminal.
View Source →Writing
Notes on machine learning, LLMs, and applied AI engineering.
RAG Retrieval Optimization: Precision, Not Recall@K
An LLM's answer is set by the precision and order of its context, not by Recall@K. Retrieving more documents, badly ordered, makes RAG answers worse.
Reduce LLM API Costs: Input Tokens Dominate the Bill
An LLM bill is a product of a few factors, and input tokens re-read on every call dominate. The biggest wins are lossless discounts, not shaved output.
AI Agent Design Patterns: Why Self-Critique Loops Fail
Self-correction degrades GPT-4 on GSM8K from 95.5 to 89.0. The agent patterns that work all share one thing: an oracle the model cannot fake.
AI Data Pipelines: Why Embeddings Break Your Audit Trail
ETL fails loudly; AI pipelines fail silently. Chunking and embedding destroy what you can audit, so contracts, quality gates and lineage must run upstream.
Let's work together.
Open to conversations about AI engineering roles, agentic systems, and collaborations, in Sydney or remote. If you're hiring or building something ambitious with LLMs, let's talk. I generally reply within 24 hours.