On-Device AI for Robotics and Physical Systems
Three talks on agentic models, memory, and vector retrieval running on edge hardware for robotics, embedded systems, and autonomous devices.
- When
- Fri, September 4, 2026 · 18:00–21:00 JST
- Where
- Tokyo, Japan · In person
- Region
- Kanto (Tokyo)
- Organizer
- Tokyo AI
- Language
- EN
- Source
- Luma
Summary
Three talks on running AI agents on edge-class hardware, aimed at engineers working on robotics, embedded systems, and applied machine learning. The evening walks up the stack layer by layer: the model, the memory, and the retrieval component, with the shared constraint that everything has to work on the device itself rather than behind a cloud API.
Ramin Hasani (Co-founder and CEO, Liquid AI) opens with late interaction models and what building an agentic model for edge devices actually requires. Stefania Druga (Staff Research Scientist, Sakana AI RSI Lab) follows on agentic memory: what an agent should retain across sessions and where that memory should live when there is no cloud to fall back on. Ewa Szyszka (DevRel Engineer, Qdrant) closes with agentic retrieval, covering how vector search fits inside constrained IoT hardware and autonomous systems.
Doors open at 18:00, talks run 18:30 to 20:00, and the last hour is networking. Foundry Labs K.K. and Qdrant support the event.
About the community
The largest international AI community in Japan, with over 5,000 members based mainly in Tokyo: engineers, researchers, investors, product managers, and corporate innovation leaders. It runs more than 80 events a year featuring 300+ speakers from startups, enterprises, and academia, and works to connect the people building AI in Japan with the global ecosystem. Sessions are run in English and are open to newcomers as well as regulars.
#on-device-ai#edge-ai#robotics#embedded-systems#agentic-ai#vector-search#machine-learning