I am an Applied AI Engineer in Sydney. I work on source-backed RAG, local inference, agent tools, workflow automation, and the software underneath them.

I began with Java and Spring systems at IBM, moved closer to product ownership in Sydney startup teams, and now own internal AI systems at Pretian Squared from discovery through production support.

What matters to me is whether a system can show its sources, explain why it refused, recover when a dependency fails, and still make sense to the next engineer who has to maintain it.

Portfolio | Resume | LinkedIn | Technical writing | Email

An evaluated RAG system over 25 Australian Information Security Manual PDFs. It combines control-aware chunking, hybrid pgvector/BM25 retrieval, cross-encoder reranking, calibrated refusal guardrails, and a live evidence pipeline. On a fixed 100-question benchmark, faithfulness improved from 0.6834 to 0.8351 and answer relevancy from 0.7216 to 0.9078.

Live system | Architecture and evaluation

An MCP-native verification layer for AI-written code. Specialist review agents inspect hallucination, specification drift, and weak verification; an evidence graph and judge layer deduplicate findings; and a deterministic patch tournament applies candidate repairs in a temporary workspace before they can win.

Live system | Architecture and benchmark

A browser-native AI development environment combining WebContainers, Monaco, XTerm.js, and a 3.8B-parameter Phi-3 model through WebLLM and WebGPU. Node.js execution and model inference run locally without a backend, with model and workspace persistence in IndexedDB.

Live system | Build note

A TypeScript toolkit distributed through npm, a CLI, and an MCP server. It exposes 115+ curated prompts, 15 MCP tools, and reusable workflow chains, backed by 147 automated checks in GitHub Actions.

npm package | Build note

April 2026 - Present

Own internal AI systems from discovery through production support. Current work includes an internal AI operations platform, a source-backed RAG assistant, local model infrastructure on an ASUS NVIDIA-enabled server, and Microsoft Graph-to-Attio CRM automation with deterministic controls and human review.

2025 - 2026

Delivered full-stack CRM, CMS, and productivity features at City Quokka and eden.pm using Next.js, TypeScript, Supabase/PostgreSQL, Express.js, AWS Lambda, DynamoDB, Docker, and Vercel. Worked directly with founders and owned features from schema design to deployment.

May 2022 - December 2024

Spent 2 years 8 months delivering enterprise banking and automotive applications. Analysed requirements, designed microservice and integration flows, developed Java and Spring Boot services and REST APIs, implemented persistence and SQL changes, diagnosed production failures, and supported CI/CD and deployments across AWS, Azure, and Linux environments.

• AI: Python, FastAPI, RAG, evaluation design, pgvector, BM25, cross-encoder reranking, vLLM, LiteLLM, WebLLM, MCP.
• Application: TypeScript, Next.js, React, Node.js, Java, Spring Boot, REST APIs, SQL, PostgreSQL, MongoDB.
• Delivery: Docker, AWS, Azure, Linux, GitHub Actions, Jenkins, Azure Pipelines, Vercel, Hugging Face Spaces.

• Master of Artificial Intelligence candidate at the University of Technology Sydney, GPA 6.75/7.0 (WAM 87), graduating December 2026.
• Building internal AI products at Pretian Squared and publishing architecture, evaluation, and behind-the-build notes.
• Open to full-time applied AI engineering and software engineering roles from December 2026.

For the broader archive, explore the project portfolio, AI engineering notes, and public learning systems.