Hi! My name is Guilherme Duarte Marmerola, and I’m a Data Scientist from Brazil.
I’m passionate about machine learning, and how it empowers us to solve tough problems in business and science.
I had the fortune of discovering ML in my early undergraduate years, totally by chance, during an internship. Since then, I’ve fallen in love with it, and had a lot of fun building cool advanced analytics initiatives in many industries, such as media, mining, insurance, retail, consumer goods, banking and real estate.
I hold a special interest about some specific topics:
Reinforcement learning and bayesian inference: particurlarly, I’m interested in the exploration-exploitation dilemma, and how to solve it with (bayesian) models that capture uncertainty. I’ve devoted a lot of time to the study of bandit problems, and how to use them as frameworks to fundamental problems in industry (pricing, resource allocation, recommendation, and others).
Dimensionality reduction and clustering: how can we visualize and interpret structure in our data, aiming to explain it to business stakeholders? Can we do better than simple K-Means clustering and PCA? What similarity measures, algorithms and representations would make this possible?
Causal Inference: in industry, a sizeable chunk of projects start when someone in your organization asks for your help answering what-if questions such as:
- what would happen if we lowered prices?
- if we recommended product A instead of B, would our revenue grow?
- if we sent e-mails to our customers instead of calling them, what would happen to our conversion rate?
All of these questions, no matter how varied, can be solved with causal inference methods. How can we go beyond Randomized Controlled Trials and A/B tests and extract value from observational data in these contexts?
Management, and data science in a business context: doing data science in industry is far from trivial. We need to balance time for experimentation with deadlines, data is oftentimes messy and stakeholders may not have a clear understanding of our work and challenges. How can you navigate this and deliver real value from ML, while maintaining the best people motivated?