dAIEDGE Innovation Days

Plenary Session PS01: Intelligence at the Frontier

Room Ballroom

• PS01-1 Overview Presentation. 09:00 – 09:15
• PS01-2 Edge AI for manufacturing: real-world robotics applications from Malta. 09:15 – 10:00 · Laurent Filipozzi, STMicroelectronics, Malta.
• Project Pitch Presentations. 10:00 – 10:30

Session A01: Accelerating Next-Generation Edge Hardware

Room Ballroom

• A01-1 Charge-Summing Hardware Engine for Efficient Convolutional Neural Networks. Ahmed Mohey, Jacek Flak, VTT Technical Research Centre of Finland Ltd., Finland.
• A01-2 ML-Driven Latency and Area Estimation for NoC-Based CNN Accelerators. Hana Krichene, CEA, France.
• A01-3 KRIA UltraScale+ MPSoC based system for EdgeAI hardware acceleration. Laurentiu Acasandrei, Silicon Austria Labs, Austria.
• A01-4 Intelligent Mesh Lighting Powered by Predictive Edge AI. Weronika Bak, NxTech AS, Norway.

Session B01: Securing the Trustworthy Edge Stack

Room Swizzles

• B01-1 Building Trustworthy Edge AI: From Design-Space Exploration to Secure Deployment. Alessandro Biondi, Scuola Superiore Sant’Anna, Italy.
• B01-2 Full-Stack Supply-Chain Integrity for Edge AI Using Immutable Bootable Containers. Casper Van Gheluwe, imec, Belgium.
• B01-3 An Efficient Privacy-Preserving Communication Framework for IoT Sensors. Azarm Yeganehfallah, Zurich University of Applied Sciences, Switzerland.
• B01-4 Who Verifies the Verifier: Evidence-Anchored Validation in Multi-Stage LLM Pipelines. Oksana Dikareva, GradeBuilder S.L., Spain.

Session C01: Neural Optimization and Cascade Pipelines

Room Ballroom

• C01-1 Confidence-Shaped Regression Cascades for Efficient Edge Inference. Kyrylo Tymchenko, INRIA, Université Côte d’Azur, France.
• C01-2 Beyond the Current Stage: Looking Back and Ahead in Multi-Model Cascade. Ying Zheng, Université Côte d’Azur, France.
• C01-3 Smart Water Tank Energy Optimisation with Edge AI. Weronika Bak, NxTech AS, Norway.
• C01-4 QuantAM: Attention-Guided Mixed-Precision Post-Training Quantization of Vision Transformers for Edge AI. Raffaele Mineo, University of Catania, Italy.

Session D01: Federated Continuum and Orchestrated Learning

Room Swizzles

• D01-1 Presentation. TBC.
• D01-2 Collaborative Intelligence Across the Edge-Cloud Continuum: Practical Federated Distillation for Resource-Constrained Edge Environments. Gleb Radchenko, Silicon Austria Labs, Austria.
• D01-3 Agentic AI for Trustworthy Autonomous Optical Inspection at the Edge. Alberto Faro, Deepsensing Srl, Italy.
• D01-4 Presentation. TBC.

Session E01: Neuromorphic Computing and Event-Based Processing

Room Ballroom

• E01-1 ARTEMIX-NeuroForge: A Mixed-Signal Design and Simulation Environment for Spiking Neural Networks. Saeid Jamili, Sapienza University of Rome, Italy, RAME s.r.l., Italy.
• E01-2 Enabling SNN Plasticity on FPGA for Low-power System Personalization on SYNtzulu. Paola Busia, University of Cagliari, Italy.
• E01-3 Neuromorphic Optical Flow for Event-Based Visual Odometry. Filippo Samorì, University of Bologna, Italy.
• E01-4 Robust Audio-Visual Speech Recognition for Human-Robot Interaction on Neuromorphic Edge Hardware. Panagiota Moraiti, PLAIXUS IKE, Greece.

Session F01: Robust and Private Federated Learning

Room Swizzles

• F01-1 VGIA: a Verifiable Gradient Inversion Attack in Federated Learning. Francesco Diana, INRIA, Université Côte d’Azur, France.
• F01-2 Secure End-to-End Federated Aggregation over Encrypted Model Updates. Varesh Mishra, KU Leuven, Belgium.
• F01-3 Secure Explainability at the Far Edge in Federated Edge Learning. Philippe Massonet, Xavier Lessage, CETIC Research Center, Belgium.
• F01-4 OMPF-EDGE: Privacy-Preserving Real-Time Workout Analysis with Personalisation and Federated Learning on Consumer iPhones. Bohdana Savchuk, BEZRYMA GROUP LLC, Ukraine.

Plenary Session PS02: Sustainable Autonomy Continuum

Room Ballroom

• PS02-1 Keynote Presentation. 09:00 – 09:30
• PS02-2 Keynote Presentation. 09:30 – 10:00
• Project Pitch Presentations: REBECCA. 10:00 – 10:30

Session A02: Quantized Transformers and On-Chip Acceleration

Room Ballroom

• A02-1 Q-WiSE: Quantized Deep Wiener-Filter Speech Enhancement for Ultra-Low-Power Micro-Edge Processors. Mohamad Hasan Bahari, SENSIFAI, Belgium.
• A02-2 On-Device Continual and Federated Adaptation for Multi-Site Solar Irradiance Forecasting. Mateusz Piechocki, Poznan University of Technology, Poland.
• A02-3 Fitting Vision Transformer Attention on a Low-Cost SoC FPGA: A Sparse, Integer-Only Accelerator. Simon Vellas, SYPHOS, Greece.
• A02-4 ARTEMIX-PACE: Design and Implementation of Compact VLM and LLM Inference on the SoC-FPGA. Saeid Jamili, RAME s.r.l., Italy.

Session B02: Autonomous Agents and Robotic Planning

Room Swizzles

• B02-1 Small Models, Specialized Skills: Enabling AI Agents on the Edge. Christos Kozanitis, FORTH, Greece.
• B02-2 Trusted AI Agents at the Edge. Pramod Bhatotia and Teofil Bodea, Technical University of Munich, Germany.
• B02-3 Locally-Deployed, Modular LLM-based High-Level Planner for Robotic Systems. Toms Eduards Zinars, EDI, Latvia.
• B02-4 The rise of physical AI. Valeria Tomaselli, STMicroelectronics, Italy.

Session C02: Advanced Vision and Semantic Mapping

Room Ballroom

• C02-1 Embedded Open-Vocabulary Semantic Mapping: A Comparative Profiling of Foundation-Model-Based Methods. Calvin Galagain, Université Paris-Saclay, CEA List, France.
• C02-2 Extending Domain-Agnostic Semantics in Terrain Segmentation: Language-Guided to Self-Supervised. Peteris Racinskis, EDI, Latvia.
• C02-3 Automatic Pseudo-masks Generation for Real-Time Vine Segmentation on the Edge. Lucas Mohimont, Université de Reims Champagne Ardennes, France.
• C02-4 Advanced Vision Processing Trends at the Edge. Alain Pagani, DFKI, Germany.

Session D02: Adaptive Allocation and Reconfigurable Scheduling

Room Swizzles

• D02-1 Efficient Resource Allocation with Dynamic Regret Guarantees through Quantized Online Gradient Descent. Isidoor Pinillo Esquivel, INRIA, France.
• D02-2 GEAR: GPU Elastic Allocation and Reconfiguration for Zero-Downtime, Priority-Aware Sharing at the Edge. Ádám Kiss, HUN-REN-BME Cloud Applications Research Group, Hungary.
• D02-3 EdgeMan: Automated Resource Allocation for Edge Testbeds. Christos Kozanitis, FORTH, Greece.
• D02-4 EXPANSE AI – Empowering X-sector Pilots for EuropeAn and interNational Scale-up and compEtitiveness in AI. Inessa Seifert, VDI/VDE-IT Berlin, Germany.

Session E02: Efficient Multimodal and Vision Models

Room Ballroom

• E02-1 What Does Visual Token Pruning Really Save? A Hardware-Aware Study of VLM Inference. Martyna Poreba, Université Paris-Saclay, CEA, List, France.
• E02-2 Token Reduction Beyond PyTorch: Benchmarking Vision Transformer Deployment on NVIDIA Jetson. Victor Bercy, CEA, List, France.
• E02-3 From Detection to Multimodal Refinement: A Hierarchical Three-Stage Edge-to-Server Pipeline for Weapon Recognition using Large Multimodal Models. Oscar Deniz Suarez, University of Castilla-La Mancha, Spain.
• E02-4 Presentation. TBC.

Session F02: On-Device Adaptation and Edge Control

Room Swizzles

• F02-1 Enabling Efficient On-Device Model Adaptation with COTS Edge Inference Accelerators. Alessandro Capotondi, University of Modena and Reggio Emilia, Italy. Mateusz Piechocki, Poznan University of Technology, Poland.
• F02-2 Precision Provisioning in Programmable Architecture for On-Device Learning: The Hidden Cost of Synchronization. Rohit Prasad, Université Paris-Saclay, CEA, List, France.
• F02-3 See-Before-Grind: Edge AI for Vision-Guided Robotic Surface Rework. Hacı Mehmet Güzey, AdaGuzey Engineering Software Mechatronics Ltd., Sivas University of Science and Technology, Turkey.
• F02-4 Adaptive RUL Prediction for Automotive MOSFETs under Variable Operating Conditions at the Edge. Alberto Faro, Deepsensing, Italy.