Research

Differentiated Access Memories, LtRAM, and StRAM

September 2024 – Present

This is my primary research direction as a Ph.D. student, where I work under the guidance of Prof. Thierry Tambe and collaborators throughout the Computer Science and Electrical Engineering departments at Stanford.

DAM (Differentiated Access Memories) is a five-year project at Stanford University to research and define how future computers will manage memory as a heterogeneous resource with different tradeoffs. I am interested in tackling the “memory wall” problem in the AI/ML fields, where the surging size of AI models demands more on-chip memory to store intermediate results, optimizing spatial and temporal locality and improving energy efficiency. However, even as the rest of the logical circuits continue to scale down with more advanced process nodes, the density of SRAM has not been keeping pace, leading to the risk where memory may dominate the area and power budgets of future AI accelerators.

This points to a renewed interest in alternative on-chip memory devices that provide higher density and improved fJ/bit efficiency. My lab and I argue that the future of memory will evolve from a uniform address space to a heterogeneous collection of different memories optimized for different uses. Our research seeks to answer: What new memory technologies will we need, how do we compose them in systems, and what caching mechanisms do they need? How can applications and algorithms guide the composition of heterogeneous memories? How will software present and manage heterogeneous memories?

My contributions to this overarching research direction include, chronologically:

• GainSight: Given 2T/3T gain cells as SRAM alternatives with higher density but limited retention times, as well as the prospects of composing different memory devices into heterogeneous on-chip memory hierarchies, I developed GainSight, the first comprehensive, open-source framework that aligns dynamic workload lifetime profiles with memory device characteristics to generate optimal on-chip cache compositions in accelerators, enabling our lab to explore the design space of heterogeneous memory hierarchies for upcoming architecture designs.
• Position papers on the Future of Memory and Towards Memory Specialization (DIMES ‘25): I co-authored two position papers that argue for a paradigm shift in memory architectures away from uniform memory hierarchies toward specialized memory architectures that exploit application-specific access patterns and data lifetime characteristics, proposing long-term RAM (LtRAM) and short-term RAM (StRAM) as new memory classes with explicit OS support.
• OpenGCRAM: Contributing to an open-source Gain Cell memory compiler for generating GCRAM designs with customizable retention times for AI workloads.

Undergraduate-level Research

Many of my research experiences prior to Stanford were done at the University of Michigan as an undergraduate student through the U-M’s Civil and Environmental Engineering department and the University of Michigan Transportation Research Institute (UMTRI).

Road-side based cybersecurity in connected and automated vehicle systems, January 2022 – April 2024

This project encompassed my research work done under the guidance of Prof. Neda Masoud at the University of Michigan Center for Connected and Automated Transportation. The overarching objective of this research project is “to develop a holistic framework that integrates physics-based data-driven modeling and dynamic decision making under uncertainty and partial information to improve cybersecurity in connected and automated vehicles (CAV).”

UMTRI Bioscience Group, February 2022 – August 2022

I worked as a research assistant at the University of Michigan Transportation Research Institute’s Biosciences Group over the Winter, Spring, and Summer 2022 terms. My group primarily “conducts research on the biomechanics of motor vehicle occupants as it relates to occupant injuries, crash protection, and occupant accommodation.”

Publications, Preprints and Conference Presentations