PhD Students

Doctoral Fellows

Maximilian Mordig

After a Bachelor's in Physics and Master's in "Computational Science & Engineering" at EPFL, I worked for Hoffmann-La Roche for roughly two years. My PhD, supervised by Bernhard Schölkopf and Gunnar Rätsch, will focus on causal inference, representation learning and applications to biological datasets. I previously worked on Bayesian optimization, variational inference and Gaussian processes. The beauty of probabilistic modelling thrills me.

Primary Host(s):	Bernhard Schölkopf (ELLIS Institute Tübingen & Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Gunnar Rätsch (ETH Zürich)

Mattia Segù

Current state-of-the-art machine learning models often unpredictably fail when dealing with previously unseen settings, proving to be not general and robust enough for a safe deployment in life-critical applications. In the spirit of delivering safe and reliable intelligent systems, my research interests focus on solving the challenges presented by the real world by making artificial intelligence models capable of continuously learning and adapting to previously unknown scenarios.

Primary Host(s):	Luc Van Gool (ETH Zürich & KU Leuven)
Exchange Host(s):	Bernt Schiele (Max Planck Institute for Informatics & Saarland University)

Paul Wrede

My aim is to design microscale robots to revolutionize the treatment and diagnosis of diseases. Besides designing the next generation of smart microrobots I am especially interested in visualizing these tiny man-made machines in-vivo. Consequently, my research interests include microrobotics, bioimaging, and micro/nanotechnologies as well as personalized medicine.

Primary Host(s):	Metin Sitti (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Daniel Razansky (ETH Zürich)

Siyeon Jang

I am interested in tiny medical robots that can generate localized electric field stimulation. Since electric fields can affect various biological interactions, such wireless nano/micro-scale robots are expected to be one of the most promising strategies to achieve effective medical applications.

Primary Host(s):	Metin Sitti (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Mehmet Fatih Yanik (ETH Zürich)

Núria Armengol Urpí

Despite recent advances in robotics, the gap between human and robot skills remains vast. During my PhD I aim to combine the tools from machine learning and control theory in order to advance in the field of autonomous learning for robotics. I try to develop algorithms that provide robots with the ability to efficiently explore new environments and autonomously learn from diverse prior data, so that they can intelligently, reliably and robustly handle the complexities of the real world.

Primary Host(s):	Stelian Coros (ETH Zürich) Otmar Hilliges (ETH Zürich)
Exchange Host(s):	Georg Martius (University of Tübingen)

Nima Mahkam

My research interests lie in developing flexible and intelligent microrobots as a promising potential approach in medical applications. I am mainly interested in fabricating small-scale mechanisms and characterizing their properties to understand the microscale dynamics better. Analyzing these tiny robots' behavior and identifying their mechanical and biocompatible characteristics are crucial for precisely controlling these micro-mechanisms in tight environments such as the human body. Consequently, my studies aim to explore new design approaches, fabrication techniques, and control principles for intelligent soft small-scale robots, which would have a high impact in healthcare.

Primary Host(s):	Metin Sitti (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Daniel Razansky (ETH Zürich)

Daniela Macari

I am interested in exploring forms of physical intelligence inspired by materials and mechanisms found in nature and in applying these to the design and fabrication of soft robotic systems. Particularly, my research focus lies on bioinspired actuation systems replicating the versatile performance of natural muscles - so-called HASEL (hydraulically amplified self-healing electrostatic) actuators. This new class of electrohydraulic transducers enables the creation of novel soft robotic systems with yet unmet high degrees of dexterity, agility and adaptability to their environments, making them well-suited for unstructured and dynamic tasks.

Primary Host(s):	Christoph M. Keplinger (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Stelian Coros (ETH Zürich)

Yasunori Toshimitsu

I like making, controlling, and playing with robots. Through my research, I hope to create robotic systems that can act in the chaotic and beautiful everyday environments that we live in, to hopefully make our lives a little bit easier. Currently, I believe that the introduction and exploitation of compliant structures are an important key to approach that problem. I hope to interact with a wide range of technologies, from the material to control level in my research.

Primary Host(s):	Robert Katzschmann (ETH Zürich)
Exchange Host(s):	Christoph M. Keplinger (Max Planck Institute for Intelligent Systems)

Jasan Zughaibi

Magnetically actuated continuum robots show promise for a variety of medical applications. Their high flexibility allows them to perform minimum invasive surgeries inside the human body. However, their modeling and control are particularly challenging due to nonlinear material behavior and their high degrees of freedom. During my PhD, I aim to develop intelligent learning control algorithms which allow the reliable deployment of these systems for high-precision medical interventions.

Primary Host(s):	Bradley Nelson (ETH Zürich)
Exchange Host(s):	Michael Muehlebach (Max Planck Institute for Intelligent Systems)

Hao Ma

My research interests lie in the intersection of machine learning and control theory. I try to build a new framework to explain stability, convergence, robustness, etc. in some machine learning algorithms. Moreover, I want to bridge the gap between theory and practice in machine learning.

Primary Host(s):	Michael Muehlebach (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Melanie Zeilinger (ETH Zürich)

Zhiyu He

My research interests lie in optimization, learning, and control of dynamical systems. Specifically, I am curious about the fundamental limits of and the interplay between model-based/physics-guided and model-free/data-driven control approaches. I aim to use these approaches to achieve optimal, autonomous, online, and distributed decision-making for engineering systems, particularly power systems and robotics.

Primary Host(s):	Florian Dörfler (ETH Zürich)
Exchange Host(s):	Michael Muehlebach (Max Planck Institute for Intelligent Systems)

Malte Prinzler

Hi I’m Malte. My research interest lies on reconstructing photorealistic 3D models of human heads with affordable capturing hardware. Recent research has made remarkable progress in creating high-quality reconstructions based on extensive camera setups. Democratizing such technology through lowering the capture costs would enable everybody to create their own counterpart in the virtual world and could spark a true revolution of how humans communicate with each other.

Primary Host(s):	Justus Thies (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Otmar Hilliges (ETH Zürich) Siyu Tang (ETH Zürich)

Artur Grigorev

My area of research is modelling the human body, in terms of both shape and appearance, in 3D space. I am particularly interested in generative models for the human body as well as in reconstruction and modelling of clothing. Currently I am working on reconstruction and modelling of loose garments, such as skirts and dresses, as I think that is the area most lacking in current approaches for creating human avatars.

Primary Host(s):	Otmar Hilliges (ETH Zürich)
Exchange Host(s):	Michael J. Black (Max Planck Institute for Intelligent Systems)

Liang Zhang

My research focuses on the intersection between optimization and machine learning. In particular, I try to establish the theoretical guarantees of machine learning algorithms, including their convergence properties and generalization behaviors. For example, I previously worked on the convergence analyses of variance reduction methods for structured nonconvex stochastic optimization that are common in many machine learning applications such as GANs, meta-learning, and reinforcement learning. Currently, I am working on differentially private stochastic optimization due to the increasing demand for data privacy in the industry. In addition to that, I am also interested in understanding the generalization of nonconvex optimization problems.

Primary Host(s):	Niao He (ETH Zürich)
Exchange Host(s):	Michael Muehlebach (Max Planck Institute for Intelligent Systems)

Marco Bagatella

I am broadly interested in reinforcement learning, and in the intersection between classical planning and learning-based methods. Other topics I like exploring span from model-based approaches to representation learning in high-dimensional environments. My research is largely motivated by the distant goal of enabling safe and efficient learning of complex behaviors in the real world, possibly from scratch.

Primary Host(s):	Georg Martius (University of Tübingen)
Exchange Host(s):	Andreas Krause (ETH Zürich)

Andreas Opedal

My research focuses on natural language processing. I am particularly interested in designing systems that have a deep semantic and syntactic understanding of natural language, which proves useful for improving robustness to distribution shifts, interpretability and performing more complex reasoning-based tasks. To that end, I am a strong believer in neuro-symbolic methods, formal language theory and causal inference.

Primary Host(s):	Mrinmaya Sachan (ETH Zürich)
Exchange Host(s):	Bernhard Schölkopf (ELLIS Institute Tübingen & Max Planck Institute for Intelligent Systems)

Florian Dorner

With the more and more widespread deployment of Al systems in the real world, they are likely going to have larger and larger impacts on society, including accidents, political destabilization and other negative externalities. To increase developers' accountability for negative societal impacts of their Al systems, my research focuses on a) a better theoretical understanding of these impacts, b) approaches to monitor and empirically demonstrate them, c) building tools developers can use to mitigate potential risks their Al systems might pose, and d) strategies for policymakers to align Al developers' incentives with society.

Primary Host(s):	Moritz Hardt (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Fanny Yang (ETH Zürich)

Vivian Nastl

My research interests lie in the area of machine learning. In particular, I am interested to enhance the understanding of algorithmic decisions in social context. In the last few years, more and more social and economic authorities have started to base their decisions on machine learning methods. Hence, it is fundamental to understand and quantify the influence such algorithmic decisions have on individuals and society as a whole. To that point, I am dealing with different concepts for causal relationships in social settings and developing suitable methods to properly infer these causal relationships from experimental and observational data.

Primary Host(s):	Moritz Hardt (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Nicolai Meinshausen (ETH Zürich)

Markos Diomataris

Considering what humans are able to understand about other humans it becomes apparent that our perception goes far beyond (while surely leveraging) our geometric aspect and is able to infer semantic information such as goals, (inter)actions, intentions etc. My goal is to bridge the gap between geometry and semantics so that machines can truly understand humans and therefore more intelligently learn from, interact with and assist them.

Primary Host(s):	Otmar Hilliges (Max Planck Fellow) (Max Planck Institute for Intelligent Systems) Michael J. Black (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Stelian Coros (ETH Zürich)

Jiduan Wu

I'm constantly fascinated by problems towards a better theoretical foundation of algorithms in reinforcement learning and optimization, especially those with significant implications in social or control contexts. More broadly, my interest lies in understanding complex systems in the real world leveraging tools from reinforcement learning, optimization, game theory, and network science.

Primary Host(s):	Moritz Hardt (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Niao He (ETH Zürich)

Eva Remlova

My vision is to interface technology with living tissue in a highly precise and minimally invasive manner. This strategy aims to reduce the patient burden and allow for targeted, more effective therapeutic and diagnostic interventions. I am particularly interested in harnessing the fundamental materials science and bioimaging modalities to advance the functionality and navigation of small-scale robotic systems for biomedical applications.

Primary Host(s):	Daniel Razansky (ETH Zürich)
Exchange Host(s):

Vanessa Sklyarova

My research interests lie at the intersection of Computer Graphics and Computer Vision, focusing particularly on creating photorealistic head avatars. Currently I am working on physically plausible hair reconstructions. I believe that by integrating principles of physics into the structure, appearance, and simulations of hair, we can significantly elevate the overall visual quality of avatars.

Primary Host(s):	Michael J. Black (Max Planck Institute for Intelligent Systems) Justus Thies (Max Planck Institute for Intelligent Systems) Otmar Hilliges (Max Planck Fellow) (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Marc Pollefeys (ETH Zürich & Microsoft)

Anna Kerekes

My research interests include, but are not limited to Machine Learning Theory, Bayesian Inference and Optimization. My academic foundation lies in the realm of Mathematics, particularly within the field of Statistics. I have previously worked on theoretical description of inductive biases and performance of optimization algorithms.

Primary Host(s):	Gunnar Rätsch (ETH Zürich)
Exchange Host(s):	Bernhard Schölkopf (ELLIS Institute Tübingen & Max Planck Institute for Intelligent Systems)

Frederike Lübeck

My research interests span around causal structure learning and causal representation learning. I am particularly interested in learning causal mechanisms in complex dynamic systems, with applications ranging from cell biology to climate modeling. Additionally, I am interested in probabilistic modeling and its use for algorithmically-informed experimental design.

Primary Host(s):	Andreas Krause (ETH Zürich)
Exchange Host(s):	Bernhard Schölkopf (ELLIS Institute Tübingen & Max Planck Institute for Intelligent Systems)

Taiki Nakano

The lack of rigorous theoretical guarantees stands as a fundamental challenge to the widespread deployment of learning-based control algorithms—or machine learning algorithms in general—in real-world systems, especially when safety is critical. To make these powerful tools more available and reliable, my research interests are in the methodological and theoretical advancements in learning-based control algorithms and machine learning algorithms. Through my research, I hope to combine ideas from both machine learning and control theory, in order to explain and enhance the properties of these algorithms, including performance, adaptability, and efficiency.

Primary Host(s):	Florian Dörfler (ETH Zürich)
Exchange Host(s):	Michael Muehlebach (Max Planck Institute for Intelligent Systems)

Jana Egli

As robots become increasingly integrated into our daily lives, I am interested in making them safe and efficient for human-robot interaction. I am especially eager to develop humanoid devices using soft HASEL actuators. I aim to contribute to quantifying their efficiency upon system integration.

Primary Host(s):	Christoph M. Keplinger (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Robert Katzschmann (ETH Zürich)

Felix Sarnthein

I am generally fascinated by the learning dynamics of neural networks. This involves, in particular, self-supervised learning, knowledge distillation, the interplay of optimization methods and deep networks, as well as long-range and hierarchical feature learning of sequential data. I believe that understanding such topics will lead to better-informed and potentially more efficient training methods and architectures in the long term.

Primary Host(s):	Antonio Orvieto (ELLIS Institute Tübingen & Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Thomas Hofmann (ETH Zürich)

Julius Erbach

My research focuses on computer vision, specifically leveraging learned priors from large deep neural networks, such as diffusion models, large language models, and vision-language models. These models provide rich prior knowledge that can enhance a wide range of tasks, including 3D modeling, scene understanding, and computational photography. I aim to explore how these priors can be harnessed to push the boundaries of visual perception and understanding in real-world applications.

Primary Host(s):	Konrad Schindler (ETH Zürich)
Exchange Host(s):	Bernt Schiele (Max Planck Institute for Informatics & Saarland University)

Alexandros Delitzas

My research interests lie in the areas of computer vision and machine learning. While past research has primarily focused on understanding the world around us through static captures, I am interested in developing methods that can reconstruct, represent and understand the dynamic 3D world we live in.

Primary Host(s):	Marc Pollefeys (ETH Zürich & Microsoft)
Exchange Host(s):	Christian Theobalt (Max Planck Institute for Informatics)

Thomas Wimmer

My research at the intersection of vision, graphics, and geometry focuses on (dynamic) 3D scene understanding, reconstruction, and generation. My work aims to develop efficient methods for (real-time) 3D scene reconstruction from minimal inputs and create robust representations that capture the complexity and temporal coherence of dynamic scenes.

Primary Host(s):	Bernt Schiele (Max Planck Institute for Informatics & Saarland University) Christian Theobalt (Max Planck Institute for Informatics)
Exchange Host(s):	Siyu Tang (ETH Zürich)

Xueyan Li

I did my Undergrad at Imperial College London in Bioengineering and Master's degree at Cambridge University in Machine Learning. Currently, I am interested in reasoning, world models, and adaptive learning in large language models. How can position encodings and memory mechanisms allow them to adapt to long-context input? I am also interested in models that look forward to anticipate event outcomes and user responses. Can models adjust their current behavior based on their predictions of the future? How can models reason better and understand causality? In my spare time, I love cooking cuisine from all over the world, baking and going on weekend trips.

Primary Host(s):	Jonas Geiping (ELLIS Institute Tübingen, Max Planck Institute for Intelligent Systems & Tübingen Al Center)
Exchange Host(s):	Mrinmaya Sachan (ETH Zürich)

Jana Zeller

Unlike humans, current machine learning models often struggle to generalize beyond their training data, fail under distribution shifts, or rely on superficial statistical patterns rather than capturing the underlying structures of data. To address these challenges, my research focuses on learning robust representations that generalize effectively across both distribution shifts and modalities. My overarching goal is to develop algorithms capable of adapting to diverse and challenging scenarios.

Primary Host(s):	Wieland Brendel (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Ryan Cotterell (ETH Zürich)

Bahri Batuhan Bilecen

My current research focuses on making generative deep networks more controllable, specifically in video and 3D-domains. I previously worked on human-centric stylization and generation tasks, as well as inverse problems in image reconstruction.

Primary Host(s):	Siyu Tang (ETH Zürich)
Exchange Host(s):	Bernt Schiele (Max Planck Institute for Informatics & Saarland University)

Shlomo Libo Feigin

I am interested in the theoretical foundations of ML, particularly in two complementary questions: 1) How to characterize inductive biases of ML frameworks? — which often requires answering — 2) How to characterize training data beyond standard i.i.d. assumptions? I explore both questions in various settings. For example, I previously worked on finding theoretically optimal data augmentations for self-supervised representation learning and on data-efficient training of diffusion models.

Primary Host(s):	Thomas Hofmann (ETH Zürich)
Exchange Host(s):	Konstantin Rusch (ELLIS Institute Tübingen & Max Planck Institute for Intelligent Systems)

Adamya Singh Dhaker

My broad research focus is on understanding the physical principles behind nature's embodied intelligence through high-fidelity modeling of soft and biohybrid robots. The inherent flexibility and compliance of soft robots make them ideal for replicating the embodied intelligence of biological systems. I aim to develop accurate computational models that not only translate natural principles into soft robotic design but also offer insights into biological systems through first-principles modeling.

Primary Host(s):	Janneke Schwaner (Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Robert Katzschmann (ETH Zürich)

Philipp Nazari

The world is full of interesting problems waiting to be solved. My current research therefore focuses on the theory of geometric deep learning, as well as its applications in the sciences. In particular, I aim to develop mathematically grounded machine learning models that draw inspiration from our physical universe.

Primary Host(s):	Konstantin Rusch (ELLIS Institute Tübingen & Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Fanny Yang (ETH Zürich)

Patrik Wolf

My research focuses on test-time training and preference alignment, often in the context of large language models. I am particularly interested in developing a theoretical understanding of these techniques and in characterizing their potential to improve model performance. Ultimately, I aim to leverage these insights to design robust and adaptive learning algorithms.

Primary Host(s):	Celestine Mendler-Dünner (ELLIS Institute Tübingen & Max Planck Institute for Intelligent Systems)
Exchange Host(s):	Andreas Krause (ETH Zürich)

Associated Fellows

PhD Students

Doctoral Fellows

Associated Fellows

ETH

MPI

MPI

MPI

MPI

Alumni