Project Engineer and Developer tinkering with AI, .NET, React & Rust
Passionate about user privacy. Oddly fascinated by file converters.
Often code from a treadmill.
MCSE-certified Systems Administrator turned Software Developer with 15+ years bridging enterprise IT and modern development. As a Project Engineer, I help manage infrastructure for 500+ employees by day while building production software that solves real business problems and pain points. My unique position allows me to identify inefficiencies and create integrations that drive real efficiency gains - from unifying disparate ERP systems into a single interface to automating warehouse operations with barcode scanners. Currently exploring the intersection of Rust, WebAssembly, and enterprise .NET while maintaining a steadfast commitment to user privacy. Whether it's replacing a $6k/year backup solution, breaking down data silos between departments, or building zero-knowledge file converters - I just love to build cool things with pretty obsessive attention to detail.
Rust and its interop capabilities: WebAssembly for browsers and FFI for .NET integration
Started while learning Rust and wanting to benchmark it against .NET - a simple file converter seemed like the perfect test case. The real challenge was getting Rust compiled to WebAssembly running in the browser, which led to this privacy-focused HEIC to JPG converter that processes images 100% client-side. Built this because HEIC conversion is a constant pain point in professional life when dealing with iPhone photos. Your photos never leave your device, and it's hosted via Cloudflare Workers for global performance.
Started this after hitting GitHub Desktop's auto-generate commit message paywall - thought it would be cool to build a tool that could use any AI model, including offline ones, to generate commit messages. What began as a simple alternative evolved into a full-featured CLI tool that respects your choice of AI provider. Cross-platform AI-powered Git commit message generator supporting multiple models (OpenAI, Anthropic, Google, Groq, local). Features instant model switching via aliases (@fast, @smart), secure encrypted API key storage, cost tracking, and automatic provider parameter detection. Built with .NET 9.0 as a single executable with no dependencies.
Production multi-agent conversational AI system built on CQRS architecture implementing Anthropic's latest research patterns from their Building Effective Agents, Multi-Agent Research System, and Contextual Retrieval articles. Features 10+ specialized agents including Web Research Agent (Brave Search API + custom Playwright scraper), SharePoint Research, Planning, Evaluation, Synthesis, Citation, and Org Chart agents orchestrated for complex queries. Multi-provider AI architecture supports Azure OpenAI, Google Vertex AI, and custom REST endpoints. Advanced RAG system ingests SharePoint documents with contextual chunking while Web Research Agent provides real-time internet access with domain filtering and intelligent content extraction. Unique whimsical thinking traces engage users during processing ("Spelunking through documents", "Orchestrating insights"). Dual database architecture (PostgreSQL with pgvector for embeddings + MS SQL Server) powers conversation sharing with expiration controls, message ratings, and real-time PowerBI dashboards with token-level observability tracking usage per user/message/agent/workflow across all providers.
Unified internal web application that sits on top of multiple disparate systems (ABM ERP, AutoCAD, Strumis, PowerBI) consolidating all company data into one accessible React/TypeScript interface, enabling self-service information for users like workshop supervisors who otherwise couldn't access these systems. Drives a sophisticated Bill of Materials system with live links between AutoCAD drawings and the ERP - when drawings complete approval workflows, procurement can bulk order directly into the ERP system, then fulfill via my custom warehouse picking app. Replaces legacy Access databases while providing enhanced functionality including multi-location inventory tracking, procurement automation, and comprehensive digital forms. Built with feature-based architecture using TanStack Router, React Query for server state, and SignalR for real-time updates. Features the AI Assistant frontend connecting to the Multi-Agent AI backend, with MSAL authentication providing role-based access control for 200+ users across all departments.
Enterprise Android warehouse management system with integrated barcode scanner support, built for handheld devices with a sophisticated hybrid offline/online architecture. Handles complete warehouse operations including inventory control, pick order fulfillment, stock transfers, and purchase order receiving. Built with Kotlin and Jetpack Compose following Clean Architecture/MVVM patterns with reactive flows throughout. Features Room DB with SQLite WAL mode for performance, gRPC for blazing-fast syncs of hundreds of thousands of ERP rows enabling true offline-first operation while maintaining real-time responsiveness when connected to the network. Microsoft SSO integration with intelligent sync strategies that adapt to network conditions. Optimized for 2GB RAM devices with intelligent batch processing and memory management. Connects seamlessly with the Integration Hub's BOM system - when drawings are approved and orders placed, warehouse staff use this app to pick and fulfill orders, completing the entire workflow from design to delivery.
Clean Architecture .NET 8 file conversion service supporting 90+ formats via Microsoft Graph API. Built as a learning project after completing courses on design patterns, SOLID, and testing. Features automatic converter discovery via reflection, multi-page document splitting (PDFs to individual JPGs), and comprehensive testing with 82% coverage (191 unit + 220 integration tests). Includes RESTful API, Console CLI, and Blazor Web UI.
Enterprise backup solution for Autodesk Construction Cloud/BIM360 in production since May 2022. Replaces a $6k AUD/year commercial solution while being 5x faster. Currently running nightly backups of 170+ projects (225GB) in 8 hours. Built with clean architecture featuring a reusable ApiClient library, fluent API design, parallel downloading with configurable concurrency, and comprehensive retry policies with exponential backoff. OAuth 2.0 two-legged authentication.
Started when I saw an industry leader charging $3k for 200 Chromebooks with mediocre software. Thought "how hard could it be to get a live updating list of tabs and a screenshot from the Chromebook into a browser?" Built it during school holidays, grew to 2K AUD MRR through word of mouth alone - teachers loved it. Also had Windows version working (C# and .NET) in a test pilot school but focused on ChromeOS first. Python handled auto updating and deployment services. Deployed across 10-15 SA schools. Shut down in 2020 after getting distracted by crypto/trading instead of focusing on the business.