Bayesian Neural Networks with Ferroelectric Memory Field-Effect Transistors (FeMFETs)
Artificial Intelligence (AI) increasingly powers safety-critical systems that demand robust, energy-efficient computation, often in environments marked by data scarcity and uncertainty. However, conventional AI approaches struggle to quantify confidence in their predictions, making them prone to unreliable or unsafe decisions.
This thesis contributes to the emerging field of Bayesian electronics, which exploits the intrinsic randomness of novel nanodevices to perform on-device Bayesian computation. By directly encoding probability distributions at the hardware level, these devices naturally enable uncertainty estimation while reducing computational overhead compared to traditional deterministic architectures.
Previous studies have demonstrated the promise of memristors for Bayesian inference. However, their limited endurance and high programming energy pose significant obstacles for on-chip learning applications.
This thesis proposes the use of ferroelectric memory field-effect transistors (FeMFETs)—which offer nondestructive readout and high endurance—as a promising alternative for implementing Bayesian neural networks.
Adaptive and explainable Video Anomaly Detection
Video Anomaly Detection (VAD) aims to automatically identify unusual events in video that deviate from normal patterns. Existing methods often rely on One-Class or Weakly Supervised learning: the former uses only normal data for training, while the latter leverages video-level labels. Recent advances in Vision-Language Models (VLMs) and Large Language Models (LLMs) have improved both the performance and explainability of VAD systems. Despite progress on public benchmarks, challenges remain. Most methods are limited to a single domain, leading to performance drops when applied to new datasets with different anomaly definitions. Additionally, they assume all training data is available upfront, which is unrealistic for real-world deployment where models must adapt to new data over time. Few approaches explore multimodal adaptation using natural language rules to define normal and abnormal events, offering a more intuitive and flexible way to update VAD systems without needing new video samples.
This PhD research aims to develop adaptable Video Anomaly Detection methods capable of handling new domains or anomaly types using few video examples and/or textual rules.
The main lines of research will be the following:
• Cross-Domain Adaptation in VAD: improving robustness against domain gaps through Few-Shot adaptation;
• Continual Learning in VAD: continually enriching the model to deal with new types of anomalies;
• Multimodal Few-Shot Learning: facilitating the model adaptation process through rules in natural language.
Development and Characterization of Terahertz Source Matrices Co-integrated in Silicon and III-V Photonics Technology
The terahertz (THz) range (0.1–10 THz) is increasingly exploited for imaging and spectroscopy (e.g. security scanning, medical diagnostics, non-destructive testing) because many materials are transparent to THz radiation and have unique spectral signatures. However, existing sources struggle to offer both high power and wide tunability: electronic sources (diodes, QCLs) deliver milliwatts but over narrow bands, while photonic emitters (photomixers in III–V semiconductors) are tunable across broad bands but emit only microwatts. This thesis aims to overcome these limitations by developing an integrated matrix of THz sources. The approach is based on photomixing two 1.55 µm lasers in III–V photodiodes to generate a phase-coherent THz current coupled to THz antennas.
Initially, the PhD student will experimentally investigate an existing 16-element THz antenna array (STYX project) CEA-CTReg/DNAQ: setting up the test bench, measuring phase coherence, optical coupling, radiation lobes, and constructive interference. These experiments will provide a scientific foundation for the subsequent design of an integrated photonic array on silicon. The student will simulate the photonic architecture (couplers, waveguides, phase modulators, Si/III–V transitions) synchronizing multiple InGaAs photodiodes. Prototyping will include the fabrication of silicon photonic circuits (CEA-LETI) and THz photodiodes/antennas in InP (III-V Lab or, to be confirmed, Heinrich-Hertz-Institut of the Fraunhofer—HHI), followed by their hybrid integration (bonding, alignment).
This thesis will also rely on close collaboration with the IMS laboratory (Bordeaux), which is nationally and internationally recognized for its expertise in silicon photonics and THz systems. IMS will provide complementary expertise in optical modeling, electromagnetic simulation, and experimental characterization, reinforcing the multidisciplinary strength of the project.
Finally, the ultimate goal of this thesis is to develop a proof-of-concept demonstrator with a few phase-locked THz emitters (e.g. 4–16) will be produced and characterized, showing enhanced beam directivity and output power thanks to constructive interference. This demonstration will pave the way for large-scale THz source arrays with significantly improved range and penetration for advanced THz imaging systems.
AI-Driven Network Management with Large Language Models LLMs
The increasing complexity of heterogeneous networks (satellite, 5G, IoT, TSN) requires an evolution in network management. Intent-Based Networking (IBN), while advanced, still faces challenges in unambiguously translating high-level intentions into technical configurations. This work proposes to overcome this limitation by leveraging Large Language Models (LLMs) as a cognitive interface for complete and reliable automation.
This thesis aims to design and develop an IBN-LLM framework to create the cognitive brain of a closed control loop on the top of an SDN architecture. The work will focus on three major challenges: 1) developing a reliable semantic translator from natural language to network configurations; 2) designing a deterministic Verification Engine (via simulations or digital twins) to prevent LLM "hallucinations"; and 3) integrating real-time analysis capabilities (RAG) for Root Cause Analysis (RCA) and the proactive generation of optimization intents.
We anticipate the design of an IBN-LLM architecture integrated with SDN controllers, along with methodologies for the formal verification of configurations. The core contribution will be the creation of an LLM-based model capable of performing RCA and generating optimization intents in real-time. The validation of the approach will be ensured by a functional prototype (PoC), whose experimental evaluation will allow for the precise measurement of performance in terms of accuracy, latency, and resilience.
Rheology and Conduction of Functional Polymers for Embedded Electronics in 3D/4D Additive Manufacturing
This PhD project, conducted on the MAPP platform (CEA-Metz), focuses on the development of additive manufacturing (3D/4D) processes for the integration of smart materials. The aim is to overcome the limitations of traditional planar electronic architectures (PCBs, wafers) integration by enabling the direct-to-shape printing of electronic functions within 3D parts performed by Fused Deposition Modeling and Paste Extrusion Modeling. The research will address functional conductive polymers, composed of an organic matrix and metallic particles, whose conduction mechanisms (direct contact, tunneling effect, ionic conduction) are governed by the percolation threshold. The study will investigate the processing of these materials, their rheological and electrical behavior, and the exploitation of their resistive, piezoresistive and piezoelectric properties to design novel sensor (3D) and actuator (4D) functions. The doctoral candidate will benefit from advanced characterization facilities and the guidance of a multidisciplinary team with expertise in additive manufacturing, materials science, and microelectronics.