Last Updated: 12/13/2024
My email: humzahm@uchicago.edu
I am a junior at UChicago, studying statistics and computer science. I really enjoy doing work at the intersection of statistics, engineering/computer science, and finance. I find some of the problems there, including my research projects listed below, to be the perfect mix of intuitive enough that anyone can understand them and their implications while also being complex, challenging, and require creativity to solve. If I'm not working I'm probably with friends or watching TV. Sometimes I'll make them watch TV with me.
Out of respect for Dr. Levy I won't share too much of the details of the current project online. My work is writing Python scripts to run LLMs on massive amounts of (financial) data and closely analyzing the output such as the conditional probabilities of tokens and other patterns. I'm really enjoying this project right now.
What I'm learning so far: Working with big data. A lot more about LLMs and NLP. That I still really enjoy research.
I spent most of the summer working on one large scale webapp. It was a website for one of the client teams to manage some large and complex email lists, including a maker-checker workflow and audit trail. It was the first Python webapp being built at Apollo ISG so I had to work with some senior team members to learn how to deploy it in the existing .NET enviroment. I also had to establish new tables in the database and figure out processes to safely interact with existing data.
I was driven on this project knowing I was delivering actual value for Apollo. Without this every single change had to be ran be ran from client teams to dev teams to IT teams, done through SQL and tested in UAT before going into production. This was 2-3 man hours each and they had on average 5 or more requests a week.
What I learned over the summer: How to navigate the bureaucracy of a large corporation to get things done. How to take a project from nothing to production. What its like to sit in an incident review meeting after you broke production while deploying.
Right now we think about markets in terms of equal chunks of calendar time - days, weeks, months, years. The project asked what if we looked at markets in terms of activity? For example, how do we split 30 years (360 calendar months) into 360 periods of equal event time? I enjoyed this project so much because we didn't have a "right" answer to this. We explored some various things:
The results were interesting though not too groundbreaking - the data becomes generally much less skewed, much more normal, and possibly also more efficent (for example, autocorrelation of returns was much lower).
What I learned: Python, Pandas, and much more about market microstructure. That I really enjoy research. That messy code leads to bad results and confusion and on my next project I need to have more discipline from day 1 and think about how everything will be scaled up after the prototyping phase.
I spent two summers at NASA Johnson Space Center through a program called the Summer Robotics Academy, for graduating seniors on select FIRST Robotics Teams. After my first summer I asked to return and they kindly said yes. The first summer I worked primarily on mechanical engineering and design tasks using CAD. The second summer I focused primarily on software related tasks. We also had a bomb squad inspired challenge my second summer - I was on the winning team.
I'm grateful for every opportunity I've ever got and I recognize that a lot of people are struggling right now. If you are a first or second year student, feel free to reach out.