Lennart Schulze
M.S. Computer Science graduate, ERP Scholar & Research Assistant at Columbia University.
New York, NY
lennart.schulze [at] columbia .edu
I am a machine learning research assistant, broadly interested in deep learning, 3D computer vision, and robotics. In particular, I am interested in building methods that can learn generalizable, multimodal, robust, explainable, and controllable neural field representations of the world that allow embodied agents to reason about their environment with common sense. Naturally, robot perception, interaction, and manipulation are applications of these methods.
Currently, I am a Graduate Research Assistant working on dynamic neural fields at Columbia’s Creative Machines Lab, advised by Prof. Hod Lipson. I have also been a Visiting Research Fellow at MIT CSAIL, where I worked on OOD robustness for language and vision models with Prof. Dylan Hadfield-Menell. Prior to that, I worked on quantum machine learning for high-energy physics at IBM Research advised by Dr. Panos Barkoutsos in collaboration with CERN. At IBM, I also held several positions in ML engineering, cloud computing, data science and AI governance, working with internal and industrial clients.
I am a Fellow of the German Academic Scholarship Foundation, of the German Academic Exchange Service (DAAD), and an ERP Scholar of the German Government.
news
Mar 19, 2024 | New preprint on arxiv. We show how latent adversarial training improves robustness in vision and large language models. |
---|---|
Jan 29, 2024 | Our paper on Dynamic neural fields for robot modeling got accepted to ICRA 2024. |
Nov 25, 2023 | Our paper on quantifiable explainability for computer vision models got accepted at NeurIPS 2023 XAIA. |
Aug 21, 2023 | Our paper on dynamic neural fields got accepted and selected for oral presentation at ICCV 2023 NeRF4ADR. |
May 29, 2023 | I am joining MIT CSAIL as a Visiting Research Fellow in the Algorithmic Alignment Group to work on OOD robustness of vision and language models. |