Adrian Vladu

I am a permanent researcher at CNRS, based in Paris at the Institut de Recherche en Informatique Fondamentale, Université Paris Cité.

I received my PhD from MIT Math in 2017, followed by a postdoctoral position at Boston University.

My research focuses on various aspects of convex and non-convex optimization. I enjoy combining tools from continuous optimization, convex geometry, and linear algebra in order to obtain improved algorithms for classical discrete problems. I am also interested in deep learning, particularly in designing theoretically-grounded training methods that scale efficiently with limited computational resources.

Recently, I have been focused on developing simple, implementable algorithms with provable guarantees, using strong theory to guide the design of practical methods. I aim to show that general techniques like interior point methods can be refined with minimal modifications to achieve state-of-the-art results for fundamental problems.

My research is currently supported by CNRS and the French National Research Agency.