Embedding methods are widely used but poorly understood. Better theory can explain when they reveal real structure and when they create it.
Lars Chen
Lars Chen is a Ph.D. student in Computer Science at the Hertie Institute for AI and an IMPRS-IS scholar. In his work, he develops theory that unifies neighbour embedding methods with diffusion-based approaches to dimensionality reduction, aiming to uncover what these methods preserve about the structure of high-dimensional data. He completed a master’s degree in Computational Neuroscience & AI at the BCCN-Berlin, where he worked on representation learning and deep causal models for neuroimaging, and was a Research Engineer in AI/ML at NYU Langone Health. His research interests include dimensionality reduction, geometry in machine learning, and mechanistic interpretability.
Embedding methods are widely used but poorly understood. Better theory can explain when they reveal real structure and when they create it.
Present Positions And Title
PhD Student
Research Group
Email Address
Career
| Period | Institution | Role |
|---|---|---|
| Since 08/2026 | Hertie Institute for AI in Brain Health, University of Tübingen | PhD Student |
| Since 2026 | NYU Langone Health | Research Consultant |
| 2025 – 2026 | NYU Langone Health | Research Engineer |
| 2022 – 2024 | Institute of Theoretical Biology, Humboldt University of Berlin | Graduate Research Assistant |
| 2019 – 2021 | Epic Systems | Technical Solutions Engineer |
Academic Education
| Year | Degree | Institution | Field of Study |
|---|---|---|---|
| 2024 | M.Sc. | Technische Universität Berlin | Computational Neuroscience |
| 2019 | B.S., B.A. | University of North Carolina at Chapel Hill | Biomedical Engineering; Mathematics |