I am a Ph.D. candidate at TU Berlin's ML Group, advised by Prof. Klaus-Robert Müller and Shinichi Nakajima. I am also a research associate at the Berlin Institute for the Foundations of Learning and Data (BIFOLD). Previously, I worked at SAP AI Research, advised by Rahul G. Krishnan, Tassilo Klein, and Moin Nabi. My research focuses on generative models for multiple modalities (such as vision and language) and probabilistic modeling. I published several peer-reviewed papers, hold one patent, and earned a Master's degree from TU Berlin.

Research Overview

I presented (preliminary) research at the Berlin Machine Learning Meetup in 2020 (video).

We represent unimodal and multimodal structure in different latent variables tied by dependencies.

We prove an elementary shortcoming of a popular multimodal VAE.

We use graph structure from an additional modality to guide the learning of an image classifier.

We use a 3D image model for data augmentation.

We develop a drone delivery system that handles allocation and execution for delivery tasks.

We conduct 34 three-hour experiments and analyze the data for emotion classification using deep learning.

We use uncertainty measures to guide the active learning of a surrogate model.

We optimize last-mile deliveries using Java.

Design inspired by Jon Barron