About Me
Welcome! I'm Dr. Andre Guimaraes, a Sr. Data Scientist at Michelin, North America, with roots in both Baton Rouge and Rio de Janeiro. Physics, programming, and data science captivate me, blending into my career and hobbies.
My academic path began at UERJ in Brazil and concluded with a PhD from LSU, where I also engaged with the LIGO Scientific Collaboration under Prof. Gabriela Gonzalez. Overcoming language and cultural barriers, I thrived in academia, even serving as president of the Society of Physics Students at SHSU.
My journey into physics was sparked by a childhood gift, Simon Singh's "Big Bang," and programming became a passion with my first job at Open Switch in Rio. Music, another love, saw me jamming in college bands.
Currently, at Michelin, I focus on Business Intelligence, forecasting, and implementing Large Language Model (LLM) and Natural Language Processing (NLP) tools for various internal applications. This role allows me to blend my expertise in programming, machine learning, and data analysis to drive innovation and strategic decision-making.
Travel is my avenue for growth, having explored the U.S., Europe, Japan, and Brazil with loved ones. I strive for a harmonious professional and personal life, aiming to apply my skills in meaningful work while pursuing my dream of world travel. Mentored by great minds like Einstein and Feynman, and supported by influential professors and bosses, I am grateful for the guidance on my journey.
Enjoy discovering more about me here on my site!
Showcased Projects and Roles
LIGO Scientific Collaboration
Leading of Research Projects developing Machine Learning tools for detector noise assessment
Liaising between LIGO Detector Characterization and Low Latency groups.
Liaising between LIGO and Vigor Detector Characterization groups.
Attending and Presenting at various conferences and meetings
Commercial Aircraft Trajectory Weather and Fuel Analysis
In this project with Ruth Mac-Leod, Dumindu de Silva, Hannah Solomon, and Noah Thompson made for the Erdös Institute Data Science Bootcamp, we analyzed thousands of different flights throughout the US thanks to the OpenSky Network team that gave us access to their database.
We used that data as well as IEM weather dataset to train a predictive model of aircraft fuel consumption based on weather conditions. We're still working on improving the project with more flights and different locations.
Risk Analysis and Profit Optimization for LendingClub Data
In this independent project I used the published Lending Club data to perform an overall EDA, seen on this Tableau Dashboard, as well as train a model to predict the probability of a loaner defaulting on their loans.
The model achieved an overall ~70% prediction accuracy, and I discovered that, by setting different thresholds on the model's default-probability there could be a ~100% increase in profitability from loans, but weeding out the right proportion of "Dangerous" loaners.
