I am a research scientist at Google Brain interested in fundamental problems in machine learning and artificial intelligence.
Before that I co-founded PriceHubble and was a doctoral student in the Learning and Adaptive Systems Group at ETH Zurich, supervised by Andreas Krause. In my PhD studies, I investigated coresets - small summaries of large data sets with theoretical guarantees - and other sampling methods for large-scale machine learning.
Our paper Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations was accepted to ICML 2019 in Long Beach.
The paper challenges a variety of assumptions in the unsupervised learning of disentangled representations and poses several research questions that we have investigated in follow-up work:
Our paper High-Fidelity Image Generation With Fewer Labels was accepted to ICML 2019 in Long Beach.
This paper investigates how to train the state-of-the-art generative model BigGAN using substantially fewer labels - for example using 10x less labels or even without any labels at all.
We will be showing a demo of the newly released Google Research Football Environment at the Google AI booth during ICML 2019 in Long Beach (Tuesday and Wednesday at 3.30-4pm).
Google Research Football is a novel RL environment where agents aim to master the world’s most popular sport—football. Modeled after popular football video games, the Football Environment provides a physics based 3D football simulation where agents control either one or all football players on their team, learn how to pass between them, and manage to overcome their opponent’s defense in order to score goals.