: D

  • Home

Hello

Shuli Jiang (蒋书礼)


 
I am fifth year Ph.D. student at the Robotics Institute ⊆ School of Computer Science ⊆ Carnegie Mellon University.
I am fortunate to be advised by Professor Gauri Joshi. I also work closely with Professor Steven Wu and Professor Pranay Sharma (who will join IIT Bombay in Jan.2025). Prior to becoming a Ph.D. student, I obtained my B.S. and M.S. in Computer Science at Carnegie Mellon University. Previously, I also worked with Professor Artur Dubrawski at the Auton Lab, affiliated with the Robotics Institute.
Job Hunting
I am looking for industry research positions. I am interested in topics on differential privacy, distributed optimization, Federated Learning (FL), LLM security and efficient training of LLM! Please let me know if you have openings! :)
About Me
My research interests lie broadly in theory and applications of federated learning, communication-efficient distributed learning algorithms, differential privacy and security issues of large foundation models.
I am currently working around three topics around secure and privacy-preseving machine learning:
  • Distributed algorithms for federated learning under limited communication cost
  • Differentially private learning algorithms that ensure the privacy of user data
  • Security vulnerabilities of foundation models, including large language models (LLMs)
  • Previously, I also worked on outlier detection, (distributed) sketching algorithms, multi-view clustering, etc.
    Contact
    Email: [first name] + j [AT] andrew.cmu.edu

    Publications

    (αβ: alphabetical order, **: contribution order)


    Preprints


  • (**) Shuli Jiang, Swanand Kadhe, Yi Zhou, Farhan Ahmed, Ling Cai, Nathalie Baracaldo
    Turning Generative Models Degenerate: The Power of Data Poisoning Attacks
    [arXiv]

  • Workshop Proceedings:


  • (**) Shuli Jiang, Swanand Kadhe, Yi Zhou, Ling Cai, Nathalie Baracaldo
    Forcing Generative Models to Degenerate Ones: The Power of Data Poisoning Attacks
    NeurIPS 2023 Workshop on Backdoors in Deep Learning - The Good, the Bad, and the Ugly (Best Poster Award)
    [arXiv]

  • Conference / Journal Proceedings:


  • (**) Shuli Jiang, Qiuyi (Richard) Zhang, Gauri Joshi
    Optimized Tradeoffs for Private Majority Ensembling
    (To Appear) Transactions on Machine Learning Research (TMLR 2024)

  • (**) Shuli Jiang, Pranay Sharma, Gauri Joshi
    Correlation Aware Sparsified Mean Estimation Using Random Projection
    The Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS 2023)
    [arXiv] [Code]

  • (**) Shuli Jiang, Robson Leonardo Ferreira Cordeiro, Leman Akoglu
    D.MCA: Outlier Detection with Explicit Micro-Cluster Assignments
    The Twenty-second IEEE International Conference on Data Mining (ICDM 2022)
    [Link] [arXiv] [Code]

  • (αβ) Shuli Jiang, Hai Thanh Pham, David P. Woodruff, Qiuyi (Richard) Zhang
    Optimal Sketching for Trace Estimation
    The Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS 2021 Spotlight)
    [Link] [arXiv] [Code]

  • (αβ) Shuli Jiang, Dongyu Li, Irene Mengze Li, Arvind V. Mahankali, David P. Woodruff
    Streaming and Distributed Algorithms for Robust Column Subset Selection
    The Thirty-eighth International Conference on Machine Learning (ICML 2021)
    [Link] [arXiv] [Code]

  • (**) Bohan Zhang, Dana Van Aken, Justin Wang, Tao Dai, Shuli Jiang, Jacky Lao, Siyuan Sheng, Andrew Pavlo, Geoffrey J. Gordon
    A Demonstration of the OtterTune Automatic Database Management System Tuning Service
    The VLDB Endowment, Vol. 11, No. 12, 2018
    [Link] [Code]


  • Technical Reports:


  • Shuli Jiang
    Communication Efficient and Differentially Private Optimization
    Ph.D. Thesis Proposal, 2024
    [Link]

  • (αβ) Theresa Gebert, Shuli Jiang, Jiaxian Sheng.
    Characterizing Allegheny County Opioid Overdoses with an Interactive Data Explorer and Synthetic Prediction Tool
    HackAuton Best Show Prize, 2018
    [arXiv][Code]

  • Shuli Jiang
    Deep Multi-view Clustering Using Local Similarity Graphs
    Master Thesis, 2020
    Advisor: Prof. Artur Dubrawski
    [Link]
  • Patent

  • Inventors: Shuli Jiang, Swanand Kadhe, Yi Zhou, Ling Cai, Nathalie Baracaldo
    Title: A System and Method to Defend Against Data Poisoning Attacks Targeting Generative LLMs
    Reference number: P202303734US01
    Filed by IBM Research
    April 2024
  • Work Experience

  • Google Research (Laser, Foresight)
    Student Researcher
    Advisor: Walid Krichene, Nicolas Mayoraz
    May 2024 ~ August 2024, Mountain View, CA, USA
    Focus: Differential privacy, Recommender systems
    Design differentially private learning algorithms specifically for training models (Factorization Machine) for ads prediction and recommender sys- tems. One project is on maximizing the privacy-utility trade-offs by mak- ing use of public features in the dataset. The other project is on privatizing an MCMC-sampling algorithm to train Factorization Machines.

  • IBM Research (Almaden)
    Research Summer Intern (AI Security and Privacy Solutions)
    Advisor: Swanand Kadhe, Manager: Nathalie Baracaldo
    May 2023 ~ August 2023, Almaden, CA, USA
    Focus: Large Language Model (LLM) security
    Investigate security vulnerabilities of large language models (LLMs) in terms of data poisoning attacks targeting natural language generation (NLG) tasks, including text summarization, text completion, table-to-text generation, etc. Design and develop defense strategies to counter-attack those types of security threats to LLMs.

  • Morgan Stanley
    Technology Analyst (Application Development)
    June 2018 ~ August 2018, New York City, NY, USA
    Focus: Outlier detection system
    Develop a data quality management system which collects real-time trading data from multiple source databases, detects potential anomalies to ensure data quality and visualizes anomalous data.

  • PreSenso Ltd.
    Software Engineer
    June 2017 - August 2017, Haifa, Israel
    Focus: Outlier detection benchmark
    Develope an anomaly detection benchmark for evaluating and comparing the perfor- mances of different anomaly detection algorithms on various patterns of anomalies.
  • Talks

  • CMU CyLab Partners Conference
    Topic: Differentially Private Incremental Gradient (IG) Methods with Public Data
    September 2024

  • NSF CPS Frontier Annual Review Lightening Talk (3-min)
    Topic: Distributed Vector Mean Estimation
    February 2024

  • AI-EDGE Students and Postdocs gathering for AI Research and Knowledge Sharing (SPARKS)
    Topic: Federated Learning and Distributed Vector Mean Estimation
    September 2023

  • CMU Robotics Institute Ph.D. Speaking Qualifier Public Talk
    Topic: Differential Privacy and Private Majority Ensembling
    May 2023
  • Service

    Reviewer:

  • Conference: SODA 2022, SIGKDD 2023, NeurIPS 2023, ICLR 2024, AISTATS 2024, SDM 2024, ISIT 2024, NeurIPS 2024, ICLR 2025, AISTATS 2025
  • Workshop: The First Workshop on DL-Hardware Co-Design for AI Acceleration 2023, ICLR Workshop R2-FM 2024, ICML Workshop FM-Wild 2024, AutoML Workshop 2024
  • Journal: IEEE Transaction on Networking 2023, Data-centric Machine Learning Research (DMLR) 2024

  • Department Service:

  • CMU Robotics Institute Ph.D. Admission Committee 2023 and 2024
  • Teaching Assistantship

  • 16-831 Statistical Techniques in Robotics, Fall 2022
    @ Carnegie Mellon University
    Instructor: Prof. David Held.

  • 10-725 Convex Optimization, Fall 2020
    @ Carnegie Mellon University
    Instructor: Prof. Yuanzhi Li.

  • 17-214 Principles of Software Construction, Fall 2017
    @ Carnegie Mellon University
    Instructor: Prof. Charlie Garrod.
  • Awards

  • NeurIPS 2023 Scholar Award, 2023
  • IEEE ICDM 2022 Student Travel Grant ($700), 2022
  • Graduate Student Assembly/Provost Conference Travel Grant ($750), 2022
  • Carnegie Mellon University Undergraduate University Honor, 2019
  • Carnegie Mellon University Undergraduate Dean's List, 2015 -- 2019
  • HackAuton Best Show Prize, 2018
  • Carnegie Mellon University Innovation Scholar, 2017 -- 2019
  • Buncher Entrepreneurship Award ($10,000), 2017