Past Projects

Target Variable Definition Simulator

2023

The objective of algorithmic systems in the hiring process is to anticipate the qualities of a “good” employee. However, defining a “good” employee solely based on algorithms is problematic. This concept is vague, subjective, and can have significant consequences on fairness. To address this issue, I created FairTargetSim in Streamlit. It is an interactive simulator that utilizes real-world data to allow users to define target variables. Its purpose is to educate non-technical individuals involved in the process and enable all stakeholders to comprehend the fairness implications associated with defining target variables. Code is available here.

This is a project completed as part of the Turing Internship Network in collaboration with Accenture Labs, led by Professor Milo Phillips-Brown.

Looking for a Handsome Carpenter! Debiasing GPT-3 Job Advertisements

2022

Generative language models like GPT-3 have opened up new possibilities for downstream tasks. However, addressing biases in these models is often not tailored to real-world impact. In our study, we use GPT-3 to create unbiased job advertisements. I participated in a project where we compared zero-shot generated ads to real ones and evaluate prompt-engineering and fine-tuning as debiasing methods. We found that prompt-engineering offered no significant improvement in bias or realism, while fine-tuning, especially on unbiased ads, improved realism and reduces bias.

This project was completed as part of OxAi Labs team and resulted in a publication where I am a co-first author. The code is available here.

PyGame game: FlyNet

2021

I wrote a simple game about catching fruit flies in pygame to learn more about Object Oriented Programming.

Microarray Analysis Program

2019

This is the first program I ever wrote. In the October 2019 Oxford DTC course titled “Introduction to Programming,” students were given a choice of four programming tasks. I selected the task of designing and writing an image analysis program for reading DNA microarray images and generating a CSV file with signal strength values for “red” and “green” wells. I successfully created this program with limited functionality in a week. The program can accurately identify the location of the wells and provide RGB pixel values for the well centers. This programming task followed a 2 week intense course in python.