Welcome to my project portfolio.
Over my first year of university, I developed a keen interest in the world of trading and the stock market, exploring how programming can be used to enhance the process. This passion led to the creation of a MATLAB-based program for cryptocurrency analysis, as well as a group C++ project focused on testing stock investment strategies using data from over 20 Fortune 500 companies, both of which are discussed below.
I developed a program to analyze over 100 cryptocurrencies in real-time with data pulled from Coin Market Cap’s API. This tool allowed users to test various moving average strategies with a set investment amount, helping identify the most profitable combinations using historical and live data. Key features included strategy optimization, profit measurement, and an interactive user interface.
As part of a large group project, I contributed to the development of a C++ platform designed to simulate stock investment strategies on data from over 20 Fortune 500 companies. The simulator supported strategies such as set deposits, dividend reinvestment, a moving average crossover strategy, and momentum-based investing.
The project utilized object-oriented programming principles like inheritance and polymorphism, with a graphical user interface providing seamless navigation, graphical analysis, and price tracking. This collaborative effort not only showcased our technical skills but also provided a real-world context for exploring investment approaches.
In my role as a Research Assistant at the Australian Institute of Machine Learning, myself and a partner developed a Python tool to process and summarize over 260,000 medical reports. Each X-ray used to train the vision-language model (VLM) required an accompanying text description, enabling the model to understand and interpret image data. Since medical reports are often written in shorthand by doctors, the challenge was to structure the information into clear, descriptive, and accurate summaries suitable for training the VLM.
To achieve this, we utilized NVIDIA’s llama-3.1-nemotron-70b-instruct LLM to generate precise and coherent outputs. This automation not only enhanced the dataset’s quality but also streamlined the preparation of training data, ultimately advancing the model’s capability to perform AI-assisted diagnostics effectively.