Sign in Subscribe

Now

What I'm doing now.

Last update: 02 July 2025, from Nutrify HQ, Brisbane, Australia

Planning a wedding

My fianće Georgia and I are getting married at the end of the year.

Planning a wedding is fun.

It's another form of creating something.

Most weddings I see online are a bit too much for me.

But that's kind of the point too. You want the event to be large and important and in front of your family and friends. A starting signal of the importance of the journey to come.

Either way, I'm excited.

We made engagement rings for each other at a ring making class.

They're silver and custom.

If you haven't been to a jewellery making class, I'd highly recommend trying it out. You can make something in a few hours that'll last a lifetime.

KeepTrack - a new app with my brother

After placing third (technically winning because the top 4 got the same prize... ;P) in the Google Gemini Kaggle competition with my entry KeepTrack, my brother Josh and I decided to use the prize money to turn it into an actual app.

KeepTrack helps you keep track of what you own via photos.

Optionally it uses AI (Google Gemini) to extract details about what it sees and autofills the information for you.

You can store receipts, images and more.

And if needed, you can export a PDF report containing all of your items (this is helpful for insurance or personal record keeping).

You can see more about it at keeptrack.app or download it on the App Store.

Working on an end-to-end Object Detection tutorial (nearly finished)

I've been digging deep into the world of object detection.

And one of the ways I learn best is to create materials teaching the thing I'm trying to learn.

So my latest tutorial is all about object detection or detecting things in images.

In the project, we'll build Trashify 🚮, a model designed to incentivize people to clean up their local area.

The code and text are done and is open-source so feel free to read through it all.

Videos walking through the code will be uploaded to Zero to Mastery in the next couple of weeks.

A presentation slide titled "Project: Trashify 🚮" illustrates a machine learning pipeline for trash detection. It displays an annotated image for "Data" (showing a hand putting "trash" into a "bin"), a neural network diagram for the "Model", and a web application "Demo" where the model successfully identifies the "bin", "trash", and "hand" with confidence scores. The project leverages Hugging Face Datasets, Transformers, Hub/Spaces, and Gradio.
In the Trashify project, we'll use the Hugging Face ecosystem to build it all from dataset, model fine-tuning and demo sharing.

Consulting project with a conservation company to use computer vision to identify different species in the wild

Speaking of object detection...

I'm having fun building custom computer vision models for species detection to run on device (e.g. you're out in the wild with no internet connection but the model still works on your photos).

Working on Nutrify has given me some great insight into how to get models to run on small devices (Nutrify's computer vision models run entirely on the iPhone).

Either way, it's been a good project of:

  • Custom dataset creation (to suit their target problem)
  • Model fine-tuning (I'm using versions of MobileNetV4 and RT-DETRv2 as the models have to run on a phone)
  • Demo creation and testing

A very similar workflow to the object detection tutorial I've been working on.

If you're looking to build your own custom models to run in compute-constrained environments, let me know.

  • learnhuggingface.com – A tutorial website teaching the Hugging Face ecosystem to help solve different machine learning problems.
  • learnml.io – A website focused on machine learning concepts and tidbits, generally more technical posts than on this website.