Andrei Margeloiu
AI Researcher at Prometheus. Former Quant Researcher at Citadel.
PhD in Machine Learning from the University of Cambridge.
About
I’m an AI researcher at Prometheus, where we are building general AI for the physical world. Previously, I was a Quant Researcher at Citadel, where I focused on fixed income and foreign exchange.
I completed a PhD in Machine Learning at the University of Cambridge, where I developed new methods to improve neural networks for tabular data (data organized in rows and columns, like transaction records), especially when there are few examples and many features. At the University of Cambridge, I also earned an MPhil in Advanced Computer Science, taught mathematical methods, and supervised master’s projects.
In my academic research, I proposed general-purpose data augmentation methods for tabular datasets based on tabular foundation models. I also worked on regularization to reduce overfitting, feature selection, and model interpretability.
Selected research
-
Tabular Machine Learning on Small-Size and High-Dimensional Data
-
TabEBM: A Tabular Data Augmentation Method with Distinct Class-Specific Energy-Based Models
-
ProtoGate: Prototype-based Neural Networks with Global-to-local Feature Selection for Tabular Biomedical Data
-
GCondNet: A Novel Method for Improving Neural Networks on Small High-Dimensional Tabular Data
-
Weight Predictor Network with Feature Selection for Small Sample Tabular Biomedical Data