I’m Kauê, an economist and data scientist working somewhere between academic research and applied ML. I’m currently a PhD candidate in Economic Theory at FEA-USP, where my research sits at the intersection of agent-based modeling, reinforcement learning, and platform economics.
My path to economics went through computer science. I did my undergrad at ICMC-USP in São Carlos, where I first got interested in agent-based simulation as a way to study distributional policy. The master’s thesis that followed took that instinct further: I trained RL agents to solve the Lucas Tree Model and asked what happens when you replace rational expectations with learned behavior. It turns out, a lot changes.
On the industry side, I spent a few years at FIT Research Institute, first as a data scientist and then as tech lead on projects for clients like HP, NexTracker, and Flex. That meant time series forecasting, cloud point processing for digital twins, and setting up a GPU cluster from scratch. Since mid-2024 I’ve been a researcher at FIPE, supporting computational and optimization work for state governments and Petrobras, among others.
On the consulting side, I run KLM Consulting, where I take on ML and computational economics projects independently. I also co-founded Agência Cognitiva with a group of computing consultants. There we work with companies on AI maturity management, strategic consulting, and implementation of cognitive systems and AI-driven workflows.
The research I’m most excited about right now: an agent-based model of platform economies built in JAX, testing stationarity in long-memory time series using RNNs, and a paper with Humberto Laudares and Carlos Nathaniel on skin color and electoral behavior in Brazil.
What this site is
cobbdouglaz is a digital garden and technical portfolio. The name is a wordplay on the Cobb-Douglas production function, . Economists will get it immediately. If you’re not one, the garden note explains it.
The garden is where ideas live before they’re finished, which is most of the time. Notes are written for thinking, not for publishing. The projects section documents things I’ve built, with the full arc: problem, methodology, results.
Everything here is evolving.