PhD Position in Energy-Efficient Machine Learning for Wearable and Augmented Reality Systems

We have an open PhD position at the intersection of machine learning, embedded intelligence and human–computer interaction. The project will explore how learning systems can become more adaptive, efficient and context-aware, creating the foundations for the next generation of wearable and augmented reality platforms.

The research focuses on developing novel ML methods that learn from limited resources, adapt locally to users and environments, and remain computationally efficient. These models will be explored and benchmarked in software, with selected approaches translated into embedded prototypes to demonstrate real-world feasibility. The overarching goal is to bridge high-level algorithmic innovation with energy-aware hardware deployment, enabling intelligent sensor systems that act as autonomous micro-agents for perception and communication.

Candidates ideally have a background in computer science, electrical engineering, or related fields, and a strong interest in machine learning, optimization, or intelligent systems. Prior experience with embedded platforms, signal processing, or model compression is a plus. We value creativity, curiosity, and the ability to connect ideas across disciplines.

The PhD will be co-supervised by Dr Christoph Leitner and Prof Luca Benini in the Digital Circuits and Systems Group at the Integrated Systems Laboratory (D-ITET, ETH Zurich).

This position is part of the SNSF Ambizione project MiNI – Multimodal Neuromuscular Interface, in close collaboration with groups at the Scuola Superiore Sant'Anna in Pisa and Imperial College London.

Application context

MiNI explores the convergence of ML, embedded intelligence, and human sensing to develop a platform for real-time neuromuscular signal interpretation. Current AR/XR systems depend mainly on camera-based tracking, accelerometers or sEMG and are constrained by power consumption and limited robustness. MiNI introduces a multimodal approach that integrates electrical (sEMG), mechanical (ultrasound), and complementary sensor data for high-fidelity movement estimation, enabling natural, efficient human–machine interaction.

Research focus and methodology

The PhD research will establish the algorithmic and system-level foundations for intelligent, energy-aware sensing nodes capable of learning and adaptation at the edge. Core directions include:

• Model foundations: Pretraining strategies, foundation models and domain adaptation for ultrasound and complementary modalities.
• Sensor fusion: Multimodal frameworks that integrate heterogeneous data streams for robust, efficient inference.
• Efficiency and adaptation: Lightweight, real-time models using quantization, compression, and adaptive execution.
• Hardware–algorithm co-design: Cross-layer optimization aligning ML algorithms with embedded hardware.
• Distributed and agentic embedded AI: Networks of autonomous micro-agents for cooperative sensing and learning.

Job description

Profile

The Digital Circuits and Systems Group at ETH Zurich focuses on advancing energy-efficient computing across the full performance spectrum, from ultra-low-power IoT nodes to high-performance systems. Its research targets an order-of-magnitude improvement in efficiency through parallelization, near-sensor processing, and heterogeneous architectures with specialized accelerators.