International Workshop on Networked AI Systems

Schedule

---

• 09:20 – 09:25: Opening Remarks

---

• 09:25 – 10:15: Keynote Speaker – Prof. Eiman Kanjo, Nottingham Trent University
• Title: Tiny Physical AI: Why Intelligence, Hardware and Software Must Co-Evolve
• Abstract: Physical AI systems must sense, reason and act under severe constraints: scarce energy, intermittent or absent connectivity, strict privacy requirements, limited compute, and the physics of a body moving through the world. Yet today’s development stacks remain fragmented. Emerging novel hardware and low-power electronics communities, are making important progress in energy efficiency, but often with limited attention to system integration metrics. At the same time, systems builders deploy intelligence on conventional hardware, often unaware of what emerging substrates could enable. Algorithms are still too often benchmarked in isolation from the bodies, devices and environments that must run them. Years of deploying intelligence across radically different embodiments expose a hard truth: hardware, software and physical form factor are inseparable co-designers of capability. A drone, for example, does not need a smaller model alone. It needs heterogeneous compute, accelerators, sensing, control loops, mechanical movement, robust software, adaptive communication and sometimes collaborative learning. Optimising one layer in isolation reshapes the constraints and possibilities of the others. This talk argues for a whole-stack co-design methodology for Tiny Physical AI. Drawing on deployments from instrumented street furniture and wearables to drones, cyber-dogs and agricultural sensing platforms, the talk shows what changes when the stack is designed together. The question is no longer whether we can run AI on a tiny device. It is what substrate, what architecture and what body this intelligence actually needs.
• Biography: Eiman Kanjo is Professor of Pervasive Computing and TinyML and Head of the Smart Sensing Lab at Nottingham Trent University. She is an Honorary Provost Visiting Professor at Imperial College London, Director of the EPSRC TinyML Network, Director of the EPSRC Green+ Network, and Director of the upcoming TinyML UK Network. She is also a co-lead on the EPSRC PROSENSING project. Professor Kanjo is the Turing University Network Academic Lead for Nottingham Trent University at the Alan Turing Institute and received the Turing Network Development Award in 2022. She serves on the EPSRC ICT Strategic Advisory Team and has secured research funding from EPSRC, DCMS, Innovate UK, the EU, DSTL, ERDF, MoD and the National Lottery Community Fund. She was named one of the Top 50 Women in Engineering by the Women’s Engineering Society and received the EdgeAI Foundation Outstanding Educator Award in 2025. Earlier in her career, she held research roles at the University of Cambridge and the University of Nottingham. Her research focuses on TinyML, edge AI, smart sensing and decentralised intelligent systems, with applications across health, defence, agriculture, cities and the environment. She advocates for energy-efficient, privacy-preserving AI systems that can operate close to people, infrastructure and the physical world.
• 10:15 – 10:30: What Do Neighbors Know? Open-World Semantic Inference Attack on Intermediate Representations
Bangjie Sun, Sean Rui Xiang Tan (National University of Singapore); Rui Xiao (Shanghai University of Finance and Economics); Mun Choon Chan (National University of Singapore); Jun Han (KAIST)

---

• 10:30 – 11:00: Coffee Break

---

• 11:00 – 11:15: GreenCam: Solar-Powered Smart Camera for Traffic Condition Monitoring
Chengbin Lei, Zimo Ma, Rui Tan (Nanyang Technological University)
• 11:15 – 11:30: Depth Priors-informed Purification Defense for Car-Borne LiDAR Vehicle Detection
Yihan Xu, Zimo Ma (Nanyang Technological University); Qun Song, Jianping Wang (City University of Hong Kong); Rui Tan (Nanyang Technological University)
• 11:30 – 11:45: See Less, Think Less, Act Fast: Optimal Budgeting of Data and Compute for the Networked AI Continuum
Jonas Schulz, Ricardo J. B. Pousa (TU Dresden); Patrick Seeling (Central Michigan University); Frank H. P. Fitzek (TU Dresden)
• 11:45 – 12:00: A Simulation Driven Evaluation of Three-tier Hierarchical Inference Learning Framework
Adarsh Prasad Behera (KTH Royal Institute of Technology); Jaya Prakash Varma Champati (University of Victoria); James Gross (KTH Royal Institute of Technology)

---

• 12:00 – 12:05: Closing Remarks