Project Scoping Developer in Glasgow, United Kingdom

Researched and wrote an article focusing on recent explainability approaches in the domain of natural language processing. It argues that the future of computing is interpretable. The published article is popular among the machine learning community, with more than 500 reads per month.

A React and Next.js web game that was designed to help people get to know each other by answering questions of varying levels of intimacy, i.e., spiciness. The game was developed as a Next.js app and deployed using the Vercel ecosystem that includes detailed analytics about usage.

Explored neural language models' limits and advantages by creating one to re-create my own writing. I published a blog post illustrating the performance and limitations of my model, including detailed steps on how to reproduce the work.

A web-based game that asks people to guess which of their friends wrote which tweet. I developed the game using Node.js, React, and the Twitter API and deployed it on my server. The game also displays a leaderboard with user scores that is connected to a Google Sheet with a Google Web API.

A research project that applies interpretability techniques to an artificial intelligence agent developed using reinforcement learning (RL). My team and I worked on training custom RL agents using TensorFlow and PyTorch. Also, we later inspected whether such agents had learned about their environment and simulated humans that acted within them.

A time series library to perform data analysis and train machine learning models on time series data. It was developed as part of the Kaggle ASHRAE Great Energy Challenge. I created the library using Python, scikit-learn, LightGBM, NumPy, and pandas.

Led the development of making the charts in the Twitter web library accessible to visually impaired users. By leveraging the Web Audio API, we enabled users to listen to charts by translating numbers to pitched sounds and written chart descriptions.