Rohan Kadekodi

Rohan Kadekodi

Postdoctoral Scholar

University of Washington

CSE 358

185 E Stevens Way NE

Seattle, WA 98195

(412) 623-9509

About

I am a postdoc in the Paul G. Allen School of Computer Science & Engineering at the University of Washington, working with Prof. Baris Kasikci. My research focuses on building efficient systems infrastructure for modern storage and memory tecnologies. I work on building systems that are able to achieve the best performance out of modern and heterogeneous memory technologies such as CXL-attached memory, GPU memory, as well as byte addressable storage for improving the efficiency of datacenter applications as well as machine learning systems.

Prior to joining UW, I completed my PhD at the University of Texas at Austin under the guidance of Prof. Vijay Chidambaram, where I developed innovative solutions for persistent memory file systems and key-value stores. My doctoral research focused on transparently achieving high performance for datacenter applications on modern byte-addressable storage technologies.

Research Interests

Tiered Memory Byte Addressable Storage File and storage systems Key-value stores Virtualization Distributed systems

Publications

Peer-reviewed Publications

Y. Gu, I. Neal, J. Xu, SC Lee, A. Said, M. Haydar, J. Van Geffen, Rohan Kadekodi, B. Kasikci. Scalable and Accurate Application-Level Crash-Consistency Testing via Representative Testing.

Proceedings of the ACM on Programming Languages 9 (OOPSLA1), 477-506, 2025

Molly Jane Nicholas, Nicolai Marquardt, Michel Pahud, Nathalie Riche, Hugo Romat, Christopher Collins, David Ledo, Rohan Kadekodi, Badrish Chandramouli and Ken Hinckley. Escapement: A Tool for Interactive Prototyping with Video via Sensor-Mediated Abstraction of Time.

In Proceedings of the CHI Conference on Human Factors in Computing Systems (CHI 2023)

Rohan Kadekodi, Saurabh Kadekodi, Soujanya Ponnapalli, Harshad Shirwadkar, Gregory R. Ganger, Aasheesh Kolli and Vijay Chidambaram. WineFS: a hugepage-aware file system for persistent memory that ages gracefully.

In Proceedings of the 28th ACM Symposium on Operating Systems Principles, 2021. (SOSP 2021)

Rohan Kadekodi, Se Kwon Lee, Sanidhya Kashyap, Taesoo Kim, Aasheesh Kolli and Vijay Chidambaram. SplitFS: Reducing Software Overhead in File Systems for Persistent Memory.

In Proceedings of the 27th ACM Symposium on Operating Systems Principles, pp. 494-508. ACM, 2019. (SOSP 2019)

Suhas Jayaram Subramanya, Devvrit, Harsha Simhadri, Ravishankar Krishanswamy, Rohan Kadekodi. RandNSG: Billion Point Nearest Neighbor Search on a Single Node.

Proceedings of the 33rd Annual Conference on Neural Information Processing Systems. (NeurIPS 2019)

Pandian Raju, Rohan Kadekodi, Vijay Chidambaram, Ittai Abraham. PebblesDB: Building Key-Value Stores using Fragmented Log-Structured Merge Trees.

Proceedings of the 15th Symposium of Operating Systems Principles, pp. 497-514. ACM, 2017. (SOSP 2017)

Jayashree Mohan, Rohan Kadekodi, Vijay Chidambaram. Analyzing IO Amplification in Linux File Systems (Poster).

Proceedings of the 8th ACM The Eighth SIGOPS AsiaPacific Workshop on Systems, Sep 2017. Best Poster Award. (ApSys 2017)

ArXiv Preprints

Rohan Kadekodi*, Z. Jin*, K. Kamahori, Y. Gu, S. Khatiri, N. Bayindirli, S. Gorbunov, B. Kasikci. DualTune: Decoupled Fine-Tuning for On-Device Agentic Systems.

arXiv preprint arXiv:2510.00229, 2025

Y. Gu*, Rohan Kadekodi*, H. Nguyen, K. Kamahori, Y. Liu, B. Kasikci. ConsumerBench: Benchmarking Generative AI Applications on End-User Devices.

arXiv preprint arXiv:2506.17538, 2025

K. Zhu, T. Tang, Q. Xu, Y. Gu, Z. Zeng, Rohan Kadekodi, L. Zhao, A. Li, et al. Tactic: Adaptive sparse attention with clustering and distribution fitting for long-context llms.

arXiv preprint arXiv:2502.12216, 2025

CY Lin, K Kamahori, Y Liu, X Shi, M Kashyap, Y Gu, R Shao, Z Ye, K Zhu, Rohan Kadekodi, et al. TeleRAG: Efficient retrieval-augmented generation inference with lookahead retrieval.

arXiv preprint arXiv:2502.20969, 2025

Research Projects

Fine-tuning local LLMs for agentic workflows

University of Washington (Ongoing)

Exploring fine-tuning techniques for local LLMs to improve their tool calling capabilities in agentic systems.

Key contributions: Fine-tuning, Tool calling, Agentic systems, local LLMs

Benchmarking and Improving the Performance of LLM Applications on End-User Devices via smart GPU sharing, scheduling and memory management

University of Washington (Ongoing)

Building a benchmark suite and an evaluation leaderboard for evaluating the performance of generative AI applications and agentic workflows on end-user devices. Also building a runtime system that accelerates local LLM performance for agentic workflows using smart GPU sharing, scheduling and memory management system for end-user devices.

Key contributions: Benchmark suite, performance evaluation, end-user devices, smart GPU sharing, scheduling and memory management

Tiered Memory Management with Controlled Allocation and Adaptive Migration

University of Washington (Ongoing)

Building a tiered memory system that uses smart allocation policies for achieving fine-grained hot data tracking, along with adaptive migrations for reacting to workload changes.

Key contributions: Smart allocation policies, fine-grained data tracking, adaptive migration strategies

RDFS: Enabling Remote DAX Memory-Mapping for Persistent Memory

University of Washington & UT Austin (Ongoing)

Building a distributed Persistent Memory manager for transparent scaling of single-node memory-mapped applications across a cluster.

Key contributions: Distributed memory management, transparent scaling architecture, cluster-wide memory mapping

Shared-state system for distributed interactive applications

Microsoft Research (CHI 2023)

Building a shared-state system with client-side caching and easy-to-use API for supporting distributed interactive applications in the cloud.

Key contributions: Client-side caching, distributed state management, cloud application support

WineFS: a hugepage-aware file system for persistent memory that ages gracefully

UT Austin (SOSP 2021)

A PM file system aimed at preserving hugepages for improving the performance of emerging PM applications.

Key contributions: Hugepage preservation, performance optimization for PM applications, graceful aging behavior

SplitFS: Reducing Software Overhead in File Systems for Persistent Memory

UT Austin & VMware Research (SOSP 2019)

A user-space file system aimed at improving performance of POSIX applications on persistent memory by converting reads and writes to loads and stores from user space, and passing metadata operations to the kernel.

Key contributions: User-space file system design, POSIX compatibility, metadata operation optimization

PebblesDB: Building Key-Value Stores using Fragmented Log-Structured Merge Trees

UT Austin (SOSP 2017)

A key-value store based on fragmented log-structured merge trees, which reduces IO Amplification while increasing throughput.

Key contributions: Fragmented LSM-tree design, IO amplification reduction, throughput optimization

Teaching Experience

CS360V: Virtualization

Assistant Instructor, UT Austin

Spring 2023

CS380D: Distributed Computing

Teaching Assistant, UT Austin

Spring 2022

CS378: Virtualization

Teaching Assistant, UT Austin

Fall 2019

Technical Talks

Profile Guided Memory Tiering (PRISM SRC Liaison Talk)

May 2024

Building high-performance storage systems for Persistent Memory (BigHPC Webinar)

July 2022

WineFS: A hugepage-aware file system for persistent memory that ages gracefully at Storage Analytics team (Google)

October 2021

WineFS: A hugepage-aware file system for persistent memory that ages gracefully at Symposium on Operating Systems Principles (SOSP 2021)

October 2021

SplitFS: Reducing Software Overhead in File Systems for Persistent Memory at Symposium on Operating Systems Principles (SOSP 2019)

October 2019

Accelerating POSIX applications on Persistent Memory at VMware Research

August 2018

Work Experience

Postdoctoral Scholar

University of Washington

Aug 2023 - Present

Mentor: Prof. Baris Kasikci

Research Intern

Microsoft Research, Redmond

May 2021 - Aug 2021

Mentor: Dr. Badrish Chandramouli

Research Intern

Microsoft Research, Redmond

May 2020 - Aug 2020

Mentor: Dr. Badrish Chandramouli

Research Intern

Microsoft Research, India

May 2019 - Aug 2019

Mentor: Dr. Harsha Vardhan Simhadri

Research Intern

VMware Research

May 2018 - Aug 2018

Mentor: Dr. Aasheesh Kolli

Research Intern

University of Wisconsin, Madison

Jan 2017 - July 2017

Mentors: Prof. Remzi Arpaci-Dusseau and Prof. Vijay Chidambaram