The Minds Behind the Models: Braxton Mensah
Braxton Mensah: From Day Trading to WorldQuant and AlphaNova — as a Freshman
Braxton Mensah is just finishing his freshman year at Indiana University. While most of his classmates are settling into introductory coursework, he’s already placed in the top 1% of U.S. teams at the International Quant Championship (#21 out of 2,328), advanced to the national stage solo, built 25+ systematic equity signals as a WorldQuant Research Consultant, and made submissions to AlphaNova that passed the full gauntlet of validation checks. His secret isn’t years of industry experience — it’s an unusual willingness to chase ideas wherever they lead.
His journey into quantitative finance started in high school, not with a textbook, but with options trading.
“I would spend a lot of time researching companies and trying to decide whether I thought a stock was going to go up or down. That eventually led me to options trading, which I did for about four years.”
During his first year at university, friends introduced him to day trading based on technical analysis. Watching them trade, he started wondering whether the same strategies could be tested with math and automation rather than chart patterns.
“I wanted to see if machine learning or an algorithm could trade it consistently. I started trying to automate the strategy while I was trading through a day‑trading prop firm. That was really my entry point into coding and data analysis.”
From there, he discovered quant competitions and data platforms, including the WorldQuant BRAIN challenge, which gave him a structured environment to test his ideas instead of relying only on his own trading results.
His AlphaNova strategy evolved through a cycle of experimentation and increasing skepticism. Most of his work focused on cross‑sectional signals—predicting which assets would outperform relative to others. He explored feature ranks, rolling changes, volatility adjustments, and interactions between different features. He tested more complex models, but found that regularized Ridge models often provided more stable performance, especially when blended with other models that looked at different parts of the data.
The hardest part wasn’t finding a model with a decent Sharpe ratio. It was finding one that was genuinely different from his existing signals.
“Some of my better‑looking models were basically close copies of an earlier signal. I started using correlation checks and projection methods to remove some of that overlap. I also tested models across different time periods and against shuffled targets. That process made me a lot more skeptical of a good backtest, which was probably the biggest thing I learned from the competition.”
Outside of finance, Braxton has also been building an open-source interview-prep overlay from scratch. The project captures audio, transcribes speech, sends queries to AI providers, and returns responses in real time. He built it around his own interview-prep workflow, with a focus on making it fast, useful, and practical during mock interviews.
“I like this project because it forced me to learn about areas that had nothing to do with finance. I had to figure out how a desktop app captures audio, turns speech into text, sends information to different AI providers, and then returns a useful response without taking forever. Seeing all those separate pieces finally work together was probably the most satisfying part.”
He also ended up using Cue himself to study, turning it into more than just a side project.
When his laptop couldn’t keep up with the computational demands of multiple model variations, he taught himself Google Cloud Vertex AI to offload heavier jobs to the cloud.
“I had to teach myself how to package the code into a container, launch a custom job, save the results, and make sure I was not wasting money on compute I did not need. It was frustrating to set up at first, but now I can run a bigger validation job and reproduce it in the same environment later. That has been a lot more useful than just renting the most powerful machine I can find.”
AI has been a constant research and engineering partner. Braxton’s background is stronger on the finance side, so he relies on AI to bridge the technical gap.
“Sometimes I have a financial idea but do not immediately know the cleanest way to test it. AI can help me turn that idea into an experiment, find problems in the code, or suggest a completely different way of looking at it. I still have to check whether the code works and whether the result makes sense. There have definitely been times when an AI gave me something that sounded right but failed as soon as I tested it.”
More recently, he’s been connecting AI agents to his own notes, research, code, and past results so the agent knows what he’s already tried. For him, the real value is that it lets him move from an idea to a live test much faster.
Braxton Mensah is only getting started. He’s currently in the SEO Career program’s Alternative Investments & Investment Banking track, targeting a Summer 2028 investment banking analyst role, and continuing to build his quant research skills. If his trajectory over the past year is any indication, he’ll be one to watch.
Stay tuned for more profiles from the AlphaNova community. If you’d like to be featured, reach out—we’d love to share your story.
Join the latest AlphaNova competition and see if your signals can survive the same rigorous tests that Braxton’s did.