Injection-Locked Oscillators based Liquid Neural Networks for Generative Edge Intelligence
Design analog liquid neural networks using injection-locked oscillators for energy-efficient generative edge intelligence.
Best for:Best for microelectronics or electrical engineering graduates with a strong background in analog circuit design and neuromorphic computing.
Overview
This PhD position focuses on translating MIT-developed liquid neural network (LNN) algorithms into physical circuit-level implementations for generative edge AI. Hosted within the Laboratory of Integrated Circuits for Energy Management, Sensors and Actuators at Universite Grenoble Alpes, the project aims to build ultra-low-power time-mode cells using oscillators that mimic liquid dynamics. The successful candidate will design, interconnect, and validate a silicon demonstrator to overcome the extreme parameter density and scaling challenges of traditional neuromorphic architectures. Applicants are expected to undertake circuit design, simulation, and experimental validation to pioneer a new generation of hardware-efficient AI systems.
What you'll do
- Translate liquid neural network algorithmic concepts into physical circuit-level implementations
- Develop ultra-low power time-mode cells based on oscillators that replicate liquid dynamics
- Interconnect time-mode cells into stable, recurrent circuit architectures tailored for generative AI tasks
- Design and simulate analog integrated circuits targeting edge computing applications
- Fabricate and validate a silicon demonstrator to evaluate hardware performance
- Collaborate with researchers to publish findings in peer-reviewed journals and conferences
Requirements
Degree required: Master of Science or equivalent degree in Microelectronics, Electrical Engineering, or related discipline
Degree offered: PhD
Language: English or French proficiency as per institutional standards
Instruction: English / French
- Strong background in analog or mixed-signal integrated circuit design
- Familiarity with neuromorphic computing or recurrent neural network principles
- Experience with circuit simulation and layout tools (e.g., Cadence)
- Understanding of low-power hardware design methodologies
- Hold an eligible Master's degree prior to the start date
Funding & terms
How to apply
Deadline: 2026-11-30
Interested candidates should submit their application through the official doctoral portal before the deadline. Ensure all required academic documents and contact details for references are included in the submission package.
- CV
- Cover letter
- BSc transcripts
- MSc transcripts
Contact
Create a free profile to unlock your personal match score, the full requirements, funding details and how to apply — in English or Persian.