OP.Market is a high-performance digital marketplace designed for instant, secure, and data-driven trading of 

 items across Steam, Epic Games and PlayStation. Built to eliminate the delays and risks of traditional peer-to-peer trading, OP.Market functioned as a real-time exchange, pricing assets dynamically, maintaining live inventory, and executing trades within seconds. At its core, OP.Market was the spiritual successor to 

, improving on the learnings from that venture at a far larger scale.

The platform’s architecture was designed to handle the over 900,000 registered users in the single year of operation and sustain 20,000+ transactions daily while reducing downtime.

, enabling deterministic, high-speed trade execution.  The game was manually re-classed and disassembled using industry standard tools such as ReClassEx, Ghidra and Fiddler.

, delivered a responsive, localised experience in multiple languages.  Windowing and lazy loading techniques were employed to ensure the website remained performant, despite needing to display over 200,000 unique items to the user with pagination.

: Through binary decompilation, traffic interception, and protocol replication, I created a custom game client with a 

 (10 MB) than the game it was emulating, dramatically reducing infrastructure costs and enabling massive parallelism.

This last point was a key differentiatior from the prior site 

, which only used a single instance.  Distributing our inventory across 30 accounts meant that we traded with far more users in parallel (especially during times of increased volatility), but also resulted in new engineering challenges.

: primarily to enable a more robust experience and hedge against the risk of a single account being taken offline, the inventory was evenly distributed across all our accounts.  As such, pricing algorithms would constantly work to transfer assets between accounts to ensure no single account became a critical point of failure by holding too much value.

 to ensure inventory synchronisation and prevent double-selling, enqueued trades employed atomic reservatioon logic using a fine grained locking system.  In the event of failure or transaction cancellation, reservations were immediately rolled-back, freeing up assets to be traded with other users.

 incoming trades were placed in a priority queue, where users would be assigned trading accounts with the least missing items.  Missing items were instantaneously transferred between accounts to give the illusion a unified inventory, but without the associated risk.

 in a synthetic economy, maintaining buy and sell orders for every item in circulation. The system continuously adjusted its bid–ask spreads using a blend of real-time trading data, supply/demand imbalance, and game-related news signals.

To protect against overexposure, each asset was subject to automated 

. As stock levels approached overbought thresholds, bid prices were automatically reduced or suspended, reallocating liquidity toward higher-performing items, effectively mirroring 

portfolio risk management in financial trading

Due to the sheer volume of trades performed through OP.Market, this scale created unintended natural network effects, as collectors and retail traders realised that any item could be bought and sold very quickly on OP.Market.  The platform also became a venue for 

 as players used OP.Market bids/asks to inform their decisions in player-to-player trading Discord servers and forums.  This led to the creation of the official OP.Market pricing guide that algorithmically tracked market sentiment.

Beyond technical refinement, OP.Market introduced several features that accelerated user growth and retention:

 that fueled viral user acquisition, encouraging individual users and larger content creators to organically promote the service to their audiences for a mutual benefit.

, supporting multiple languages and regional currencies, enabling rapid expansion into the South American and MENA markets.

 and real-time chat integration to sustain global operations and guide users through technical difficulties.

 via Grafana for pricing performance, latency metrics, trade success rates, and system health monitoring.

These innovations translated directly to business impact, at peak, OP.Market generated 

 and established itself as the largest, most efficient Rocket League trading platform in the world.

OP.Market was ultimately shut down after 1 year of operation when 

 removed trading from the game.  This move was largely prompted by the 

 buyout of the development studio, in which the new ownership sought to transition the game's monetisation to the 

model.  At this point, the team at OP.Market had grown to 25 talented people across business, financial modelling, product development, marketing and live support.  Thankfully, this team and the tribal knowledge acquired were brought over to 

 where we continue to work in the similar, albeit uniquely challenging 

OP.Market

Rocket League Swaps (rlswaps.com) was the first fully automated trade bot and marketplace in 

.  While the concept of trading bots in game and financial economies as a whole was not new, RLSwaps brought the idea to Rocket League's player-to-play market, bridging real-world financial automation principles with the game's virtual cosmetic ecosystem.

Much alike modern day high frequency trading (HFT) firms, RLSwaps functioned as a market maker, providing liquidity to trading in Rocket League by maintaining thousands of active buy and sell positions across every cosmetic item in the game.  The bid-ask spread was dynamically adjusted using a variety of signals; real-time trading data, game news and item fundamentals.

To maintain healthy liquidity and avoid overexposure to low-value/common assets, the system incorporated inventory risk management logic.  Individual items were targged with "overstock" thresholds, preventing the bot from accumulating too much of a single illiquid asset.  As supply approached this limit, bids were gradually lowered then ultimately removed, reducing our exposure to underperforming items.

At its peak, RLSwaps managed the largest inventory by both asset count and total market value in the Rocket League virtual economy.  This scale created natural network effcts, as collectors and traders alike knew that if an item existed, it could be found or sold to RLSwaps.  Thus, the platform became a significant venue for price discovery as players quoted RLSwaps bids/asks across Discord trading groups and other similar forums.

RLSwaps

 was a Rocket League focussed "get paid to" (GPT) earn site, where users could earn free Rocket League items by participating in surveys, watching videos and playing games.  With over 100,000 registered users, and millions of monthly requests, 

 was one of my largest projects at the time.

 was been an ongoing project since 2019 where I acted as the lead developer, working in a team small team of 5.  The project was a unique and useful experience in working with larger-scale production environments due to the challenges that come with a global active user base.

Cloudflare proxy, caching & DNS management

RL.Supply

 is a cross platform email templating client, commissioned by 

, a multinational cosmetic surgery clinic.  This desktop application, built for Windows, macOS and Linux using Electron allows Fue Perfect employees to quickly send consistent and polished appointment emails to customers.

After connecting to a company email address, the Fue Perfect Email App provides a template form, accompanied by a live email preview containing dynamically generated graphics depending on the surgery methods selected.  Once completed, the HTML formatted email can be sent, with a log kept for record-keeping purposes.

 on GitHub, with automated builds released on the 

Fue Perfect Email App

 is a Visual Studio Code formatter extension that aligns instructions, arguments and comments in MIPS assembly code.  The formatter creates neat columns, and when combined with a syntax highlighting extension and snippets creates a familiar and powerful editor, similar to that of more popular, higher level languages.

I initially created this formatter because I found myself writing a lot assembly code for a class at university.  As I liked the neatness that came with viewing the compiled instructions in hex editors and didn't want to tediously format the code myself, as any good programmer would, I took 2 hours to write a  barebones extension to automate the task.  After sending it to some classmates, I eventually published the extension on the Visual Studio Code Extension Marketplace, leading to thousands of users installing it.

This project gave me familiarity with Visual Studio Code's formatter API and allowed me to better the programming experience for myself and friends around me.

This project is completely open source and 

.  Feel free to contribute, report bugs or just check out the source code.

Assembly Formatter

 is a strict, type-safe middleware designed for Node.js applications, leveraging the robustness of 

 schemas to validate incoming request bodies, parameters, and queries. This package seamlessly integrates with 

 to provide developers with a type-safe, declarative approach to ensure data integrity and prevent invalid or malicious data from affecting their applications.

 was to provide a pretty terse API that validated and typed requests.  Traditionally, these two would be separate activities, but there's no reason why they can't be combined.  In fact, combining, the two, 

 validation with inference is the easiest way to make sure that they both stay in sync.  In other words, the types a developer uses at build time are the types they will see in production, no matter what the user inevitably ends up inputting.  Here's a relatively feature complete example validating, the request body, query and path parameters.

The library also accommodates more commonly requested features such as global/route-specific custom error handlers, allowing the developer to customise whether or how errors are logged and displayed to end users.

Initially created a personal project to reuse the same validation logic and typing I was creating across different projects, the library has since garnered a lot of traction amongst the community.  With 4,000 downloads a week and growing, it is one of my most widely used open source software publications.  This community engagement has also had the benefit of introducing new contributors to the project as users encounter issues or take it upon themselves to propose and implement new features.  If you're interested in using this project, or contributing yourself, please check it out on 

Express Zod Safe

 is an innovative VR app that streams live camera feeds from a VR headset to a website, allowing remote users to place interactive AR markers in 3D space. These markers appear in real-time in a VR user's field of view, guiding them through complex tasks with precision and ease.  Developed over 2 days as an entry into UNIHACK 2024, we definitely bit off more than we could chew technically, but it made for some very interesting learnings about the emerging technology of VR/AR.

We built InstructAR using a combination of technologies and frameworks. The VR encoding and streaming uplink is powered by Python with OpenCV and NumPy, while the AR overlay is implemented using Unity (C#) and native Meta Quest 3 APIs. The front-end of the website was developed with TypeScript using Next.js (React), and the back-end integration was achieved with Node.js running TypeScript behind Cloudflare and an NGINX reverse proxy. Each of these 4 applications communicates with each via Socket.IO except for the headset and the steaming uplink which communicate via ADB.

Throughout the development of InstructAR, we encountered several challenges:

: Ensuring low-latency, high-quality video streaming was critical. We tackled this by establishing low-latency connections via Socket.IO from the browser to the server and the server to the headset video processing uplink. Using compression and variable frame timing, we are able to stream frames within milliseconds to a remote browser anywhere in the world.

: Achieving precise alignment of AR markers in the VR environment required an intricate understanding of realtime 3D to 2D projection and vice-versa. This involves raycasting into a 3D mesh of the real world, using the headset's position, orientation, marker location on the 2D plane and corrections for rotation and lens distortion of the video livestream.

: Designing an intuitive UI for both web and VR users presented a unique challenge as InstructAR needs to make collaboration easier, not harder. This means designing a system no hidden tricks and a simple interface.

InstructAR as a project has a lot of room for growth and certainly some polish.  More specifically, the following are some immediate steps both feature and product wise to make the technology more accessible and useful to a broarder market:

Expand the functionality of InstructAR to support customisable markers, text annotations and a native voice link.

Enhance the scalability of our platform to accommodate larger user bases.

Explore partnerships with headset and wearables creators such as Meta, Apple and Rayban to broaden the application scope and platform availability of InstructAR.

In all, there is definitely the potential of InstructAR to transform remote collaboration by bringing expert knowledge to where it's needed most.

InstructAR - UNIHACK 2024

 was an inventory management system allowing users to borrow, reserve and remotely view inventory of the school Makerspace.  The Makerspace was a technology and creativity lab where students could use 3D printers, electrical tools to create Arduino projects, robots and even rockets.  The space featured a large number of tools and components students could borrow, meaning keeping track of every item became difficult.  As such, the 

 project was created; a central system to keep track of who is borrowing and using any given item from the Makerspace.

Designed with mobile at the forefront, the app is available as a Progress Web App, installable on phones, tablets, desktops and also accessible via any web browser.  This provides offline support, rapid loading and navigation through the Single Page Application design.

Admins are able to manage inventory, upload images from their phones, view and edit loans, reservations and users in order to provide an accurate digital representation of the various tools present in the Makerspace.  At the same time, students can mark items as borrowed or reserve items they take from the Makerspace to show staff and other students where items may be.  Furthermore, students also receive reminders in the form of push notifications when a loan is overdue or a reserved item becomes available.

 of this project are open-sourced on GitHub.

Makerhive

 is an annual contest open to teams from universities around the world.  Organised by Engineers Australia and sponsored by Weir Materials, the competition sees Mechanical and Mechatronics students design and build a robotics solution over 6 months.  During my exchange to Kuala Lumpur, Malaysia in 2023, along with 3 other colleagues, I competed in the 2023 Warman Design and Build Competition and ultimately placed 1st runner-up at the campus competition.

In the fictional world of Gondwana, inhabitants utilise hybrid propulsion engines powered by liquid fuel and oxidiser stored in spherical vessels. A recent near-miss incident highlighted the risks associated with mishandling these vessels, leading to the redesign of storage zones to ensure strict separation between fuel and oxidizer containers. The task for student teams was to design and build a scale demonstration system capable of autonomously collecting three fuel vessels (simulated by tennis balls) and three oxidizer vessels (simulated by squash balls) from their respective drop zones and depositing them into the correct storage silos without any contact between different types of vessels.

All teams were required to adhere to the following specifications:

 The system must fit within a 400 mm³ cube and weigh no more than 6 kg.

 The system must operate autonomously, requiring no human intervention or control during the task.

 The system must prevent any contact between fuel and oxidizer vessels at all times.

 The task must be completed within 120 seconds.

 The system must not damage or contaminate the competition track or vessels.

 are available on the Warman Design and Build Competition Website.

Our submission to the competition was an autonomous robot with a triple-jointed robotic arm.  The arm was the most complex part; requiring a powerful servo with enough torque to lift the whole weight of the arm when fully extended and loaded.  While it certainly would've been easier to use a double or single-jointed arm, to fit within the space constraints of the competition, we needed the robot to be able to fold itself into a very compact shape for transport.

Throughout the competition, my primary focus was our robots microcontroller, the brain of the robot.  We used the common Arduino Uno board with a high power shield and relays to interface with the various sensors and motors.  I worked on our sensing and actuation system; taking readings from the onboard compass and ultrasonic sensors, then translating that into actions; voltages sent to our servos and linear motors.  This included writing two libraries to interface with both the wheels and movement system more declaratively.  Instead of requiring control over each individual motor or servo just to turn or adjust the arm position, these common functions were presented as simple declarative APIs such as 

.  This enabled rapid testing and prototyping, with our time and mental effort no longer spent on writing precise voltages to specific pins.

As is the spirit of the competition, the source code for this project is available on 

Warman Design and Build Competition 2023

After configuring many servers to deploy Node.js applications, I felt the need  to streamline the process.  As such I've developed and published a collection of 

 tailored for hosting multiple Node.js servers behind Cloudflare's firewall/CDN. These configurations are designed to provide Full (strict) end-to-end encryption, ensuring secure and efficient communication between clients and servers.

 is a popular reverse proxy that routes incoming requests to their respective destination.  It also handles caching, load balancing and can even act as a mail server, making it a powerful tool for hosting a variety of apps on a single physical server.

, the following snippets can be symlinked into the configuration folder to provide the following base functionality.

 file establishes base server settings, including logging setup, TLS configurations, and the inclusion of supplementary config files.

 snippet introduces general security headers to all requests, implementing restrictive CORS policies, iFrame blocking, and tracking prevention measures.

, all HTTP connections are redirected to HTTPS, utilizing certificates for secure communication.

 file configures various request and response timeouts to 30 seconds, balancing performance and resource utilization.

 snippet ensures that essential information is correctly passed to the server within each proxy location block.

 to a local Node.js server running on port 

, the following server block can be used:

This configuration ensures that incoming requests are securely handled and efficiently proxied to the appropriate Node.js server.

By leveraging these NGINX snippets, I've aimed to create a modular and secure approach to server configuration, facilitating easier management and scalability for multiple Node.js applications. These configurations make it the default to create robust solutions that enhance both security and performance in web server management.

For a detailed overview and access to the configuration files, please visit the 

NGINX Snippets

 is an online community and trading platform dedicated to the popular vehicular soccer video game, 

. It serves as a hub for players to connect, trade items, and discuss the game. 

In addition to their website, Rocket League Garage offers a 

 that enhances the trading experience. The bot provides various features such as item price checking, quick item search, and automated shop notifications. Users can easily inquire about the current value of items, search for specific items they wish to acquire, and find information on the in-game competitive ranks of them and their friends. The 

 streamlines the trading process and helps players find the items they desire more efficiently.

 services well over 1 million users across 10,000 servers making it one of the largest and mostly widely used bots in the community.  As such, the bot was built with performance and reliability in mind, making use of Discord's REST API over their more commonly used Gateway.  It also features the ability to check a trading partner against the Rocket League scam list for user safety and trade security.  The full list of commands currently available can be seen below:

Rocket League Garage Discord Bot

Demo Demon is an overlay mod for Rocket League that provides information on your demolition stats in-game, designed to interface with 

.  The overlay provides your KD ratio for both the game and your session and a lifetime demolition count.

Configurable settings: enable or disable game kd, session kd or total demos

The Demo Demon Bakkesmod plugin is available for installation on 

Demo Demon

In late 2022, I along with 5 other fellow engineering students entered into Monash University's 2022 Robot Building Competition.  The competition spanned 2 full days and featured a largely hardware hackathon style environment.  Teams were to build a robot, capable of following a complex coloured course on the floor as fast as possible.

2 wheels, each individually driven by a motor

The setup also included many jumper wires, a breadboard, a base, lots of tape and a batter to power the robot.

With every team given identical components, construction, design and most importantly good software separated teams apart.  At the end of 2 days, 40 teams had completed a robot, although to various levels of success, with some robots struggling to follow a straight black line and with other confidently manouevring tight turns and taking coloured shortcuts.

for its wiggly movement) surpised us and the competition.  Lobster completed all 3 courses, 2 of which in the fastest time recorded in the entire competition but unfortunately struggled with 1 other course, ballooning our average time.  In all, we came 8th and had the opportunity to meet many like minded and passionate engineering students interested in robotics and Mechatronics engineering.

The software we eventually wrote and ran along with the hardware configuration we used are available, open-source on 

Robot Building Competition 2022

 was a rapidly built hackathon style utility website, designed to generate colored chat messages for the popular FPS game, 

.  In June 2021, a feature was discovered in the game that allowed users to send colored chat messages through the use of XML-like tags.  In response, I built a quick and performant Next.js React site that would generate rainbow or colored text from user input.

 was released, the feature was removed from Valorant.  The website, while dormant, is still live as a demonstration of what used to be possible.

Portfolio