Philipp Nazari
PhD Candidate at Max Planck ETH Center for Learning Systems (CLS)
I am a CLS PhD candidate in the CAMAIL group, advised by Dr. Konstantin Rusch (Max Planck Institute for Intelligent Systems / ELLIS Institute Tübingen) and Prof. Fanny Yang (ETH Zürich). I am currently interning at Google Research, working on efficient language modeling.
My research is on efficient machine learning across the full stack, from architectural design through post-training compression (quantization, pruning) to inference-time interventions like adaptive compute. I am interested in how these axes compose, where they trade off against each other, as well as the theoretical foundations.
Previously, I obtained my bachelor’s degrees in Physics and Mathematics from Heidelberg University and my master’s degree in Mathematics from ETH Zürich, where I worked in the lab of Prof. Helmut Bölcskei.
News
| Nov 18, 2026 | I have started an internship at Google Research, where I am working on efficient language modeling. |
|---|---|
| Nov 03, 2026 | Our preprint Quantizing Recurrent Language Models Is Easier Than You Think is now public! See the blog post. |
| Oct 29, 2026 | Our preprint LinQuant: Functional Invariances for Low-Precision Linear Recurrences is now public! |
| Oct 14, 2026 | Our preprint Depth Memory is now public! See the blog post. |
| Jun 25, 2026 | I am excited to give a talk on Theoretical Perspectives on Efficient Architectures in the seminar Modern Numerical Methods for Theoretical Physics at Heidelberg University. |
Latest Posts
| Oct 16, 2026 | Quantizing Recurrent Language Models Is Easier Than You Think |
|---|---|
| Oct 14, 2026 | Depth Memory |
| Jun 04, 2025 | Dual Complexity Measures for ReLU Networks |