As AI shifts toward small, efficient models running directly on devices, systems behavior becomes impossible to ignore. EdgeAI only works when models can reason about memory, latency, bandwidth, and power in the environments where they actually run. This meetup brings together people interested in that space and looks at what it means for the next phase of intelligent systems.
We’ll talk about Architecture 2.0 and how models interact with real system signals, walk through early work in ArchEval, and share a first look at QuArch, a benchmark developed by Harvard and Qualcomm that turns hardware behavior into structured reasoning tasks. The goal is to open up a conversation around systems-aware AI at the edge and what it unlocks for research, tooling, and education.
If you’re working on SLMs, EdgeAI platforms, ML-for-Systems, or agentic models that need to understand the constraints they operate under, this session is for you.
Food and drinks will be provided.