About Me

I’m currently a researcher/software engineer at Computing Infrastructure Lab of Bytedance Inc. My current work focuses on building next-generation AI serving infrastructures. Previously I was a Postdoc Researcher of Computer Science at UT Austin, working with Prof. Aditya Akella at UT Networked Systems group (UTNS). I was one of the recipients of CIFellow 2021, supported by both National Science Foundation (NSF) and Computing Research Association (CRA). During my PhD I worked on making real time data processing systems more elastic, under the supervision of Prof. Indranil Gupta at Distributed Protocols Research Group (DPRG). My current research mainly revolves around all aspects of resource-efficient cloud computing frameworks, including serverless computing and alternative system architectures. See my CV and list of publications here. Don’t hesitate to contact me if you’re interested in my experience or would like to chat about my recent research! I’m always open to opportunities for collboration and/or full-time research positions.

Education

  • [12/2021] Ph.D. in Computer Science, University of Illinois at Urbana Champaign
  • [05/2015] M.S. in Computer Science, University of Illinois at Urbana Champaign
    • Advised by Indranil Gupta.
  • [05/2013] B.S. in Math and Computer Science, University of Illinois at Urbana Champaign

Publications and Preprints

  • [Preprints] Bodun Hu, Jiamin Li, Le Xu, Myungjin Lee, Akshay Jajoo, Geon-Woo Kim, Hong Xu, Aditya Akella “BlockLLM: Multi-tenant finer-grained serving for large language models[arxiv]
    • A block-based LLM serving framework for fine-grained resource provisioning.
  • [Preprints] Le Xu, Divyanshu Saxena, Neeraja J. Yadwadkar, Aditya Akella and Indranil Gupta. “Dirigo: Self-scaling Stateful Actors For Serverless Real-time Data Processing.[arxiv]
    • A virtual actor model that helps stream processing applications to scale freely at fine granularity.
  • [EMNLP 24] Bodun Hu, Le Xu, Jeongyoon Moon, Neeraja J. Yadwadkar, Aditya Akella. “MOSEL: Inference Serving Using Dynamic Modality Selection.[link]
    • A modality-selection framework for real-time ML inference for multi-modality data.
  • [IROS 23] Peter Schafhalter, Sukrit Kalra, Le Xu, Joseph E. Gonzalez, and Ion Stoica. “Leveraging Cloud Computing to Make Autonomous Vehicles Safer.[link] [arxiv]
    • A systematic approach that improves driving safety for autonomous vehicles with the unreliable resource pool of the cloud.
  • [VLDB 22] Li Su, Xiaoming Qin, Zichao Zhang, Rui Yang, Le Xu, Indranil Gupta, Wenyuan Yu, Kai Zeng and Jingren Zhou. “Banyan: A Scoped Dataflow Engine for Graph Query Service.[link] [arxiv]
    • A scoped dataflow model for graph traversal queries that explicitly exposes concurrent execution and control of any subquery to the finest granularity.
  • [PhD Thesis] Le Xu. “Elastic techniques to handle dynamism in real-time data processing systems.[link]
  • [NSDI 21] Le Xu, Shivaram Venkataraman, Indranil Gupta, Luo Mai, and Rahul Potharaju. “Move Fast and Meet Deadlines: Fine-grained Real-time Stream Processing with Cameo.[link] [arxiv] [slides] [video]
    • A performance target-aware scheduling frameowrk that supports fine-grained operator scheduling, built for actor-based, distributed dataflow runtime.
  • [CODASPY 20] Long, Yunhui, Le Xu and Carl A. Gunter. “A Hypothesis Testing Approach to Sharing Logs with Confidence.[link]
    • An end-to-end framework that allows users to identify risks of information leakage in logs, and to release the logs with a much lower risk of exposing the sensitive attribute through log obfuscation.
  • [SoCC 18] Kalim, Faria, Le Xu, Sharanya Bathey, Richa Meherwal, and Indranil Gupta. “Henge: Intent-driven Multi-Tenant Stream Processing[link] [slides] [arxiv]
    • An intent-driven mechanism to unify user-defined performance objectives and improve cluster-wise overall satisfaction in multi-tenant stream processing system.
  • [VLDB 18] Luo Mai, Kai Zeng, Rahul Potharaju, Le Xu, Steve Suh, Shivaram Venkataraman, Paolo Costa, Terry Kim, Saravanan Muthukrishnan, Vamsi Kuppa, Sudheer Dhulipalla, and Sriram Rao. “Chi: a scalable and programmable control plane for distributed stream processing systems.[link] [slides]
    • A generalized control plane and control message design for stream processing systems that allows a wide range of functionalities being implemented and efficiently executed.
  • [EuroSys 18] Mainak Ghosh, Ashwini Raina, Le Xu, Xiaoyao Qian, Indranil Gupta, and Himanshu Gupta. “Popular is Cheaper: Curtailing Memory Costs in Interactive Analytics Engines.[link] [slides] [tech report]
    • Replication and routing strategy designed for popularity-driven workloads for interactive analytics engines.
  • [SoCC 2017] Mijung Kim, Jun Li, Haris Volos, Manish Marwah, Alexander Ulanov, Kimberly Keeton, Joseph Tucek, Lucy Cherkasova, Le Xu, and Pradeep Fernando. “Sparkle: Optimizing spark for large memory machines and analytics[link] [arxiv]
    • A shared-memory shuffle engine and off-heap memory store that optimize Spark in the scale-up setting.
  • [IC2E 2016] Le Xu, Boyang Peng, and Indranil Gupta. “Stela: Enabling stream processing systems to scale-in and scale-out on-demand.[link] [slides] [video]
    • Exploring topology-aware algorithms for migrating real time tasks to optimize distributed stream processing system throughput during cluster configuration changes.
  • [Master Thesis] Le Xu. “Stela: on-demand elasticity in distributed data stream processing systems.[link]
  • [IWCA 2015] Wenting Wang, Le Xu, and Indranil Gupta. “Scale Up vs. Scale Out in Cloud Storage and Graph Processing Systems. “Stela: Enabling stream processing systems to scale-in and scale-out on-demand.[link] [slides]
    • Constructing cluster’s linear pricing model for both scale up and scale out cluster based on pricing scheme provided by major cloud providers.

Industrial Experiences

  • [June 2019 - Aug 2019]: Research Intern – Alibaba Damo Academy (Data Analytics and Intelligence Lab)
    • Building hierarchical actor-based framework for distributed graph querying service.
    • Supervisor: Kai Zeng
  • [May 2017 - Aug 2017]: Research Intern – Microsoft (Cloud and Information Services Lab)
    • Building a control layer inside of a real-time stream processing engine for flexible and efficient online monitoring and re-configuration.
    • Supervisor: Kai Zeng, Rahul Potharaju
  • [May 2016 - Aug 2016]: Research Intern – Hewlett-Parkard Labs (Software Analytics Group)
    • Conducting Spark performance analysis for micro-benchmark and machine learning applications
    • Supervisor: Mijung Kim, Jun Li

Teaching

[05/2018, 05/2020]: Teaching Assistant: Cloud Computing Capstone (Coursera)

[01/2015, 09/2019]: Teaching Assistant: Distributed System (CS 425)

[01/2015, 01/2016]: Teaching Assistant: Cloud Computing Concepts (Coursera)

[01/2016]: Teaching Assistant: Advanced Distributed Systems (CS 525)

Service and leadership

[Program Committee]: 2024 USENIX Symposium on Networked Systems Design and Implementation (NSDI 24’)

[Program Committee]: 2023 ACM/IFIP/USENIX International Middleware Conference (Middleware 23’)

Honors, Memberships, Awards

  • 2021: 2021 CRA/CCC Computing Innovation Fellows
  • 2020: Rising Stars EECS Workshop
  • 2019: SOSP 19 Travel Grant
  • 2018: OSDI 18 Travel Grant
  • 2018: SoCC 18 Travel Grant
  • 2017: SOSP 17 Travel Grant
  • 2016: David J. Kuck Outstanding M.S. Thesis Award
  • 2015: Grace Hopper Celebration Travel Fund
  • 2015: Conference Travel Grant
  • 2015: Outstanding Teaching Assistant
  • 2011: Member of PI MU EPSILON: National Math Honor Society
  • 2010: Edmund J James Scholar

Cloud-Free Zone

  • Sometimes I blah about random thoughts. ATTENTION: you are about to enter a cloud-free zone!
  • Occationally I tweet (or retweet) about things (those are not entirely cloud-free).