Edge Cases by AGI, Inc. #002 on Near-Memory Execution for LLMs with Lead Architect @NVIDIA

​A lot of the edge AI work we find interesting is either pre-print or too niche. With the Edge Cases we're creating a space for these topics.

​Every other week at AGI HQ, one researcher takes the room through a paper or a work-in-progress. On-device inference, small models, efficient training, anything that lives on constrained hardware.

​Case #002:  Mochamad Asri (Lead Architect at NVIDIA, PhD
 ECE from UT Austin) is presenting “System Architectures for near-memory execution”, how he thinks about bandwidth, latency, and end-to-end QoS for serving generative AI and LLMs.

​Limited to AI researchers and students only.

• ​One paper, one presenter, 45 minutes

• ​Off-the-record Q&A

• ​Drinks and conversation after

​Approval-only. 15 seats. If you're in, we'll send the calendar invite with the address.

Case #003: Open call.

​We're sourcing this session's paper from the community. Pitch your own work or a paper you want to lead a discussion on.

​👉 Apply here