Active Projects
Ph.D. | 2019 - Present
After finishing my Undergrad, I was offered a Ph.D.! My PhD is titled “Real World Solar Energy Power Prediction and Use Optimisation,” and focuses on developing and applying novel forecasting, planning and optimisation techniques for use with renewable energy. I build on the work done in my dissertation with an aim to improve and expand it. I want to improve our forecasting methods and investigate optimisation methods for energy storage (e.g Batteries)
Chemise | 2022 - Present
It started out with a plan to be a lightweight wrapper around Flax, a deep Learning library built for JAX, to take care of all the boiler plate. Over time it has evolved into a Keras like library that I use to train all the models in my PhD over the last 18 months. It can do a lot of fancy things thanks to the power of JAX; Model and Data parallelism; automaticity working on multiple GPUs; logging of metrics; checkpoints and much more. It supports loading data from tf-datasets (and im working on adding support for basic in memory datasets).
EUMETSAT Downloader | 2020 - Present
A tool to download satellite imagery from EUMETSAT, the European Organisation for the Exploitation of Meteorological Satellites (now that's a mouthful). The library is designed to download the and process the high res, high frequency, images produced by the high rate SEVIRI from the MSG satellites. After downloading and loading the files into pngs, they can be loaded into a number of formats for various uses.
Other stuff have done
Aleitha | 2020 - 2021
Another ML tracking library. I know, there are loads of them out there.
I decided to build my own as none of the ones out there exactly fit my workflow, and since I only use it for my Ph.D. work, I can make it as bespoke as I want.
It also is a fantastic way to play with some new tech / ideas and get better at writing code.
I stopped working on this in summer 2022 and moved over to MLflow as it does almost everything I need.
Dissertation | 2018 - 2019
My dissertation, aka individual project. The goal of the project was to develop methods for forecasting solar energy.
I developed a machine learning model in TensorFlow that used satellite weather imagers to forecast irradiance.
autoCrat | 2013 - 2015
A Google Sheets Add-on used by over 80,000 people worldwide as a document merge utility for Google Drive. Originally build by Andrew Stillman for New Visions for public schools, the charity where volunteer, the Add-on version was rebuild from the ground up and is maintained by me.
