About me
I'm Gareth, a hacker, maker and full-stack developer with a keen interest in combining software and electronics into bespoke products for myself and for clients.
Over the last 10 or so years, I've produced a range of devices, gadgets and tools with capabilities that range from environment monitoring to computer vision, AI to e-ink dashboards.
I love the process of designing, developing and building custom electronics and I am always looking for a new idea to make a reality. If you have one, please do get in touch
What I've worked on
E-ink Bus Timetable
Transport for London make all their line and stop data available via a public API, which opens up countless options for those with the knowledge of how to consume it.
A live view on the expected arrival times for an area's local buses is a great example and in May '23 a client commissioned this wall-mounted timetable to help them check their timetables as they were going out the door.
A low-power SBC fetches the latest data from TfL and updates the listings on the 7" ePaper display every 5 minutes. The display really does resemble paper and it's wide viewing angle and clarity in bright light make it a perfect option for this info-at-a-glance application.
Aviation weather wall map
While working on this project I learned just how closely amateure aviation is tied to weather conditions, with licenses to fly based on experience with wind, visibility and other factors. So how do you know what route you'll be permitted to take on your day in the plane? Enter METAR map.
Rather than look up tables of weather codes on a series of websites, this client devised an attractive wall map with integrated RGB LEDs that reflect the official aviation codes for various weather conditions. Their end users vary from individual pilots to the airfields themselves and the maps are both remarkably eye catching and informative wherever they're used.
AI object recognition
Designed as an educational aid, this device had to be portable, battery powered and incorporate a touchscreen and camera, but the main challenge came from integrating a machine learning model pre-trained on 400 million images and their descriptions and is able to caption and identify the contents of novel images with accuracy comparable to state-of-the-art computer vision systems without needing to be trained specifically for the task at hand.
The model allows the embedding of images and text into the same vector space, which makes it possible to compare the content of an image with text descriptions. In this case, this means being able to give the model a set of images then presenting it with a novel drawing or object and have it find the closest match
Offshore seaweed farm monitor
With the aim of validating the claims that offshore seaweed farming could be a source of low-carbon food, three test farms off the coast of Wales are being monitored by battery-powered Raspberry Pi boards fitted with pan & tilt cameras and 4G networking hardware.
Initially, the boards are recording and uploading video at regular intervals to monitor the status of the buoys, but the ambition is to use computer vision to carry out basic checks locally, only signalling for human attention when things appear out of the ordinary.
Bunnyball for the Playdate
The Playdate is a modern take on the retro handheld gaming consoles like Nintendo's Game Boy. Packing a surprisingly powerful processor behind a black-and-white screen more like an e-ink display than anything you'd find on a typical smart device, I couldn't resist trying my hand at game development for the tiny system.
Bunnyball is my homage to the old Game Boy game "Penguin Wars", where household pets duke it out in a dodgeball tournament across living room, garden and playground.
Available on itch.io from late 2024, Bunnyball represents my first (but certainly not last) foray into commercial game dev.
Get in touch
Whether you've got an idea you want to develop into a real device or just want to chat about something you've seen here, fill in the form below to send me an email.