PhD Position in Decentralized Resource-Constrained Machine Learning

PhD Position in Decentralized Resource-Constrained Machine Learning

The Distributed Computing (DISCO) Group is a research group at ETH Zurich, led by Prof. Dr. Roger Wattenhofer. We are interested in a variety of research topics on new and upcoming areas off the beaten paths. Our three main areas of research are machine learning, distributed systems, and theory of networks. Within these three areas, we are currently working on several projects: graph neural networks, natural language processing, algorithmic learning, fault-tolerance, blockchains, consensus, cryptocurrencies, digital money, central bank digital currency, decentralized finance, financial networks, e-democracy, voting, social networks, online analysis with delay, and theory of distributed algorithms. In our group, we work on both theory and practice: some members of our group focus on algorithms and mathematical proofs, and some on system design and building.

You will be entrusted with designing, developing, and evaluating data-driven methods, algorithms, and systems for three independent but related research directions in the eDIAMOND project, namely:

• Distributing model training and inference over a network of resource-constrained devices.
• Online, context-aware adaptation of Federated Neural Network Architectures based on the available system resources (e.g., communication, computation, energy).
• Communication-efficient knowledge exchange among networked federated large models.

These research directions allow you to gradually build your own research profile according to your interests, while remaining within the project’s goals. Each research direction is composed of a sequence of Tasks that will collectively achieve the project’s goal. For each Task, you will be responsible for the typical scientific research workflow: motivating the problem, identifying the main methodological shortcomings in the literature, designing and developing novel systems, planning and executing experiments, and reporting findings in articles to be published at top venues according to the project’s schedule. Periodic meetings and feedback will ensure the success of your degree and of the project overall.

We are looking for a new member with the following profile:

• A strong interest and motivation to contribute to the eDIAMOND project’s goals.
• Theoretical and practical experience and passion in at least one of the following topics:
• Machine Learning: Deep Learning, Federated Learning (vertical, horizontal), Split Learning, Model Selection, Knowledge Distillation, Low-Rank Adaptation (LoRA), Large Models (Language, Vision).
• Decentralized and Distributed Systems: Gossip Protocols, Consensus Algorithms, (Byzantine) Fault Tolerance, Peer-to-Peer Networks.
• Online Algorithms: Partially-Observable Markov Decision Processes (POMDPs), Cooperative and Competitive Multi-Agent Reinforcement Learning (MARL), Multi-armed Bandits, Bayesian Optimization.
• Automated Model Design and Tuning: Neural Architecture Search, Hyperparameter Optimization.
• Computer Networking: Resource-Constrained Networking (e.g., Internet of Things), Wireless and Mobile Communications (e.g., IEEE 802.11, Bluetooth, etc.), Protocols (e.g., MAC, Network, and Transport).
• Mathematics: Statistical Learning, Stochastic Processes (e.g., Percolation Theory, Queuing Theory, Age of Information), Network Calculus, Graph Theory, Convex and Non-convex Optimization, Approximation Algorithms.
• An excellent Master’s degree in Computer Science, Engineering, Mathematics, or other fields related to the project’s domain, from a reputable university, to be obtained before 1 September 2025 (depending on visa requirements).
• A strong Transcript of Records, stating the list of passed exams (particularly those related to the project and the areas above) and their grades.
• Strong programming skills (i.e., Python and main Machine Learning frameworks such as Pytorch).
• Experience with scientific writing (e.g., reports, theses, and scientific articles).
• Strong critical thinking, English communication skills (oral and written), interpersonal abilities, collaboration aptitude.
• Published peer-reviewed articles on project-related topics in reputable conferences or journals.
• Experience with network simulators such as OMNeT++ or ns-3.
• Experience with High Performance Computing (i.e., SLURM).
• Experience with Git.

Workplace

We offer

• Your career with impact: Become part of ETH Zurich, which not only supports your professional development, but also actively contributes to positive change in society.

In line with our values, ETH Zurich encourages an inclusive culture. We promote equality of opportunity, value diversity and nurture a working and learning environment in which the rights and dignity of all our staff and students are respected.

Curious? So are we.

We look forward to receiving your online application with the following documents:

• A short letter of motivation
• A CV
• The Transcripts of Records for both Bachelor and Master degrees, containing the list of courses taken and grades achieved
• Names and email addresses of three people who could provide a recommendation/reference letter
• Additional relevant documents: diplomas, certificates, theses, project reports, or published articles

The provided documents and certificates will be checked for authenticity.

The letter of motivation’s purpose is to assess your interest and motivation in pursuing research in the field of Decentralized Resource-constrained Machine Learning, and the match with the eDIAMOND project’s goals.

Questions regarding the position (not applications) should be directed to Dr. Antonio Di Maio (adimaio@ethz.ch).

Please note that we exclusively accept applications submitted through our ETH Job Portal. Applications via email or postal services will not be considered.

ETH Zurich is one of the world’s leading universities specializing in science and technology. We are renowned for our excellent education, cutting-edge fundamental research and direct transfer of new knowledge into society.