William Whitty

An innovative IoT application for True Digital Park in Thailand, a new venture offering co-working spaces and office rental to large companies. The application was built to streamline new and existing member's office meetings, with the ability to integrate other IoT devices such as air conditioning systems and facial recognition access gates, to name a few.

The architecture restrictions support live access for scaling numbers of people. Locally, events might require a sudden influx of up to 3-5,000 participants. The application was to be re-sold as a bundle of off the shelf components, and so modularity must enable reach to different markets without including bloat or unpaid systems.

I designed and implemented the full architecture stack and built approximately 85 percent of the application. The architecture was a modular micro-service structure, built on top of MolecularJS. At any time, a new configuration could be generated for a client, and only three of the total 25 services constructed were required.

In this way, the company was able to sell a slim operating platform to clients without extra bloat or features which the client did not require.

I've always been passionate about game development and felt the need to give back. I reached out to my building owners (a large co-working and office space) and offered to teach a session for a range of skill levels.

After meeting with the owners and discussing the project scope and requirements, I quickly gathered a team from my company and worked with them to outline the tasks, workshops, and presentations for the workshop.

On the day of, the event had a good turnout (around 60 people), with ages ranging widely. I gave presentations with two others in separate rooms, and visitors worked on implementing changes as presented to the small game I had prepared for the workshop.

Participants loved the event, prizes were given out for those who could achieve small tasks, and I was proud to see that it inspired excitement in some of the university students who attended. (In one case, a student figured out how to gather all the stars on the map by changing some of the code in the player file).

I wrote two custom engines, first in Lisp and then in C++: github.com/tavurth/gamesavvylisp

I utilized Three.js for interactive websites as well as browser-based 3D games. I also created an implementation of the CLOD terrain: github.com/tavurth/terrain-example

(link.springer.com/chapter/10.1007/978-3-540-28643-1_78)

I contributed to the core engine C++ source code of the Godot engine: github.com/godotengine/godot/pulls?q=author:tavurth+is:pr

I joined multiple game-jam competitions to release small prototype projects.
(https://tavurth.itch.io/)

I have longer-term personal game development projects, including:

• Dynamically loading procedural worlds

• Procedural generation of plants and trees

• Shader-based AI (texture-based data input)

• Shader-based terrain deformation

After some time in Bangkok, I wanted to push my abilities in machine learning. I've always been a fan of manipulating market data for learning purposes, as there's a large potential with a high difficulty curve.

Beginning by writing several basic machine learning models, I was led towards an evolutionary process to grow machine learning models, battle them against each other, propagate the winner, and repeat.

After a month or two of tweaking, I had achieved an effective progressive machine learning generator, spitting out a new and improved AI schema at the rate of one per 10-15 minutes running on a Raspberry Pi.

After a few months of running, these systems identified some interesting nuances in certain markets. I constructed an automated trading system around these ideas using the Godot game engine as the front end, and C++ machine learning modules as the back end.

The system performs admirably well and I now happily spend some of my free time refining and reconfiguring these models.