This iRacing LED Display project was great fun, getting hands-on with a Raspberry Pi and learning to link hardware and software. In this project I also learned about socket programming, which was new to me but very interesting.

The project consists of a client app, running on the Raspberry Pi, and a server app on my PC. The two chat with each other via sockets to make the LED Display show real-time iRacing data.

While it might look simple, this project is proof of my ability to learn, adapt and create.

This website serves as both a learning experience and a portfolio to showcase my learning.

I'm trying to work with many different technologies. Building web apps has been a great help in broadening my base knowledge.

Additionally, although it may not look it, it has been a good opportunity to learn design in Figma.

This website is universally rendered using Nuxt3 and deployed on Vercel (wow, that was easy!)

In January, I started a Python data analysis course with Purple Beard. This is my capstone project required to complete the course.

This project explores temporal changes in the sentiment of tweets in my hometown in 2020 using a RoBERTa-base model for the sentiment analysis.

I created a Selenium web scraper to collect the data as Twitter API access is now a little out of my price range...

Pandas, Numpy, Scikit, Scipy, Matplotlib & Seaborn are used for the analysis and displaying of data.

I created this scraper to acquire tweets without having access to the Twitter API. It’s published as a PyPI package that I’m updating as frequently as possible to keep up with Twitter's changes (as I write this, the scraper is broken due to Twitter adding a new auth loop. I have a fix ready, but not implemented.)

I use Seleniumwire to navigate the page while also reading the incoming requests to get the tweet data and monitoring the rate-limit.

I intend to add multithreading to increase the rate of tweet scraping as a partial fix to the rate limiter.

This project continues to be both incredibly satisfying and incredibly frustrating.

They say the kindest acts are those done with nothing expected in return. SendJoy embodies this, allowing anyone to anonymously send goodwill messages around the world.

SendJoy utilises an LLM to analyse the messages, ensuring only those containing good intent and positive sentiment are passed on to the recipient.

The web app is built on Nuxt3, interacting with a Python API for the sentiment analysis and data flow/management.

Feel free to send some love and joy to those you hold close!

This Flask API is primarily built to handle sentiment analysis for the SendJoy app.

The actual sentiment analysis section has been redacted in this cloned repo to stop bad actors from exploiting the app.

It also generates images using Pillow from the message content and automatically posts to Instagram using the Facebook API. Sensitive data is removed from the raw messages using NLP.

The API is hosted on DigitalOcean hooked up to a MongoDB Atlas instance.

Currently, this isn't public as I want to double-check for vulnerabilities before open-sourcing.

After fiddling with Retool and realising it didn't have the functionality to support my needs, I made this iOS app using Swift (another great learning opportunity).

This app allows me to control some of the SendJoy backend utilities when I'm AFK.

Messages that don't pass our benchmark for positive sentiment land in this app for manual review.

This app is also used for vetting the messages I feed into the Flask API for posting on social media.