Spring

Thursdays, Noon-1PM, Gates 459

Date	Speaker	Topic
4/2	David Tennenhouse	Welcome Back and Updates
4/9	Nestor Cuevas	LLM fine-tuning using eDRAM with 12us retention time
4/16	Philp Levis	National Science Foundation Expeditions Proposal
4/23	Shahir Abdurrahman	A New Architecture for Communication-Efficient Mixture of Expert Transformers
4/30	Marco Donato	MemSysExplorer
5/7	Sam Dayo	3D Long-Term Memories (LtRAMs) for Model-Resident Inference Architectures
5/14	Ioanna Vavelidou	Defect-Induced Silent Data Corruption
5/21	Colin Drewes	Optimized Scaleup of Static and Dynamic Multi-Chiplet Inference
5/28	David Shim, Katherine Mohr	LtRAM Integrated Main Memory
6/4	Agur Adams	Update to Memory Conservation on SmartNICs with Flow-Based Addressing
6/11	Christian Kubicka	µVLA: Designing a Multi-Chiplet 16nm SoP with Heterogeneous Memory on Package (HMoP) Enable Real-Time Physical AI on the Edge
6/18	Final Exams Week	Ice Cream Social

Winter

Thursdays, Noon-1PM, Gates 459

Date	Speaker	Topic
1/8	Winter workshop planning
1/15	Colin Drewes	Practice talk: 3D Integration Optimization with MIQP
1/22	Post-workshop discussion
1/29	Shahir Abdurrahman	Route-Once Model Architecture
2/5	Xinxin Wang	3T Gain Cell – FeMFET CIM for Accelerating Masked Attention Mechanism in Transformer for Image Segmentation
2/12	Zhiqiang Xie	Hierarchical Context Caching for Language Model Serving
2/19	Yao Hsiao	Consistency-Directed Formal Verification of Cache Coherence Protocol Implementations
2/26	Yasmine Omri	Agent Memory: System Implications and Co-Design for Long-Horizon LLMs
3/5	Chun Deng	Page Placement with Memory Fit
3/12	David Shim	Server LtRAM
3/19	Ice cream social

Autumn

Wednesdays, Noon-1PM, Gates 415

Date	Speaker	Topic
9/24	Agur Adams and David Shim	Packets not Pages
10/1	Shahir Abdurrahaman	Towards Memory Specialization: A Case for Long-Term and Short-Term RAM (DIMES practice talk)
10/8	Philip Levis	The Future of Memory: Limits and Opportunities (BigMem practice talk)
10/15	Shuhan Liu	IEDM practice talk
10/22	Avanika Narayan	Minions update
10/29	Colin Drewes	Optimized Scaleup of Static and Dynamic Multi-Chiplet Inference
11/5	Suhyeong Choi	Foundry Monolithic 3D Unlocks Large Throughput Benefits (IEDM practice talk)
11/12	Philip Levis	<a href-"https://arxiv.org/abs/2510.06189">Barbarians at the Gate: How AI is Upending Systems Research
11/19	Akhilesh Balasingam	Compiling declarative recurrences into optimized spatial accelerators
11/26	No meeting
12/3	Yao Hsiao
12/10	IEDM