Timetable & learning outcomes
Artificial intelligence (AI) and deep learning applications currently require significant computing resources, making low-latency, privacy-sensitive applications difficult to implement. Today, vast amounts of data are processed in remote high-performance computing clusters, raising security and energy efficiency concerns. A promising alternative lies in executing machine learning models directly on embedded devices, a field known as "Embedded Artificial Intelligence" (EAI).
To achieve this, highly optimised machine learning models must be paired with a new generation of lightweight, energy-efficient embedded hardware designed for executing neural networks at the edge. This requires close integration between hardware and software, leveraging advances in ferroelectric technologies, neuromorphic computing and hardware-aware AI acceleration.
The educational objectives of this summer school reflect the interdisciplinary nature of embedded artificial intelligence, focusing on:
› Fabrication, characterisation and modeling of AI-dedicated hardware
› System design, simulation and hardware/software co-design for AI applications
› Application-specific topics, including speech and text processing and AI-powered autonomous systems
The core goal of
this summer school is to explore hardware-related aspects of neural networks
while equipping participants with the necessary foundational knowledge
for their implementation. Topics covered will include:
› Introduction to neural networks and their adaptation to
hardware constraints
› Hardware-optimised AI models for embedded and edge
computing
› Ferroelectric materials for low-power AI
acceleration
› Electrical characterisation of energy-efficient AI
devices
› Logic cell design and memory architectures for AI
hardware
› System-level simulation and AI hardware exploration
› TCAD modelling and compact device simulations in 3D
layouts
› Fabrication of vertical Gate-All-Around (GAA)
transistors for AI acceleration
› AI-enhanced circuit design for next-generation
computing architectures
› Transformer architectures for speech and natural
language processing
› Hardware-based security: AI applications in device
authentication and RF signature recognition
› Ferroelectric-based architectures for
energy-efficient computing
› AI applications
in biomedical signal analysis and cancer detection
The programme is highly interactive, combining lectures, hands-on workshops, system simulations and real-time participatory discussions to maximise learning outcomes. Participants will gain both theoretical knowledge and practical skills, enabling them to design, implement, and optimise hardware-accelerated AI solutions for resource-constrained applications.
This summer
school also provides an opportunity for participants to engage with leading
experts from FVLLMONTI, FerroFutures, FIXIT and Ferro4EdgeAI,
offering a unique and interdisciplinary perspective on the future of hardware-based
artificial intelligence.
Tentative programme
Local and international experts from academia and industry will lead the plenary lecturers. A poster session will also be organised, allowing participants to present their research projects and discuss the connections with the summer school topics. Several practical sessions are also planned, during which participants will work with dedicated design software and visit the measurement laboratory.
Over the course of the programme, students learn in a variety of ways, starting with content related to technology layers, followed by circuit design and neural networks implementation, and ending with specific applications.
Please note: the schedule is presented in Central European Summer Time (CEST).
|
Monday June 23rd |
Tuesday June 24th |
Wednesday June 25th |
Thursday June 26th |
Friday June 27th |
| 11.15 – 12.00 Participant welcome Poster session - Networking |
09.00 – 10.00 Nanotechnologies and embedded AI - Introduction Jens Trommer |
09.00 – 10.00 Materials confrontation - Ferroelectric devices characterisation |
09.00 – 10.00 Spiking neural networks with memristive devices |
09.00 – 10.00 AI memory and computing performances - P. conference - Project focus - FIXIT Stefan Slesazek |
| 10.00 – 10.30 Coffee break |
10.00 - 10.30 Coffee break |
10.00 - 10.30 Coffee break |
10.00 - 10.30 Coffee break |
|
| 10.30 - 11.15 Advanced 3D fabrication techniques and process Guilhem Larrieu |
10.30 - 11.15 Multilevel switching dynamics in ferroeletric films |
10.30 - 11.15 Circuit and bitcell design Stefan Slesazeck |
10.30 - 11.15 Low energy AI - Participatory conference Deep dive into the Ferro4edge project Nicholas Barrett |
|
| 11.15 – 12.00 Heterointegration for emerging technologies |
11.15 – 12.00 FerroFutur: CEA-List: application datalogger |
11.15 – 12.00 Deepen understanding of neural networks and neuromorphic architecture |
11.15 – 12.00 AI applications - Transformers, speech, and languaget technologies |
|
| 12.00 – 13.00 Lunch |
12.00 – 13.00 Lunch | 12.00 – 13.00 Lunch | 12.00 – 13.00 Lunch |
12.00 – 13.00 Lunch |
| 13.00 - 15.00 Fourth industrial revolution and engineer sciences |
13.00 - 15.00 Hands on session: Device and circuit measurement techniques and practices (1/2) Marina Deng, Chhandak Mukherjee |
13.00 - 15.00 Hands-on session: TCAD and DTCO tools (1/2) |
13.00 – 15.00 Demo on the Ginestra tool (applied materials) |
13.00 – 15.00 Hardware/software co-optimisation for machine learning at the edge Giovanni Ansaloni |
| 15.00 – 15.15 Coffee break |
15.00 – 15.15 Coffee break |
15.00 – 15.15 Coffee break |
15.00 – 15.15 Coffee break |
|
| 15.15 – 17.15 Introduction on hardware for AI - History and prospective |
15.15 – 17.15 Hands-on session: Device and circuit measurement techniques and practices (2/2) Marina Deng, Chhandak Mukherjee |
15.15 – 17.15 Hands-on session: TCAD and DTCO tools (2/2) |
15.15 – 17.15 Physical characterisation of ferroelectirc materials and stacks |
|
| Dinner |
A certificate of participation will be awarded to students upon completion of the course.
Programme may be subject to change.