Dujian Ding smiling in graduation attire

Dujian Ding

Research Scientist · Meta Superintelligence Lab

I am a research scientist at Meta Superintelligence Lab, building the data flywheel for coding agents. My work focuses on usability and self-improvement: making agents easier to work with and enabling them to improve themselves. I contribute to the Muse model series, including Muse Spark and Muse Wonder.

I completed my PhD at the University of British Columbia (UBC), advised by Professor Laks V.S. Lakshmanan. My doctoral research focused on making AI more efficient and cost-effective through large language model routing, adaptive inference, and data and model management.

Previously, I earned bachelor's degrees from Simon Fraser University (SFU) and Zhejiang University, and spent eight months working with Professor Martin Ester on tumor detection.

I see three forms of self-reproduction—cells, capitalism, and AI—each nourishing the next.

Publications

  1. Preprint

    Towards Data and Model Management: From DBMS to DMMS

    Dujian Ding, Xiangpeng Hao, Yifei Li, Bicheng Xu, Keke Huang, Laks V.S. Lakshmanan

    Full version
  2. 2025

    ThriftLLM: On Cost-Effective Selection of Large Language Models for Classification Queries

    Keke Huang, Yimin Shi, Dujian Ding, Yifei Li, Yang Fei, Laks V.S. Lakshmanan, Xiaokui Xiao

    VLDB · 2025

  3. 2025

    Navigating the Prompt Space: Supervision Matters in CoT When Reasoning Misleads

    Xiang Zhang, Juntai Cao, Chenyu You, Dujian Ding

    ACL · 2025

  4. 2025

    BEST-Route: Adaptive LLM Routing with Test-Time Optimal Compute

    Dujian Ding, Ankur Mallick, Shaokun Zhang, Chi Wang, Daniel Madrigal, Mirian Del Carmen Hipolito Garcia, Menglin Xia, Laks V.S. Lakshmanan, Qingyun Wu, Victor Rühle

    ICML · 2025

  5. 2025

    EcoAct: Economic Agent Determines When to Register What Action

    Shaokun Zhang, Jieyu Zhang, Dujian Ding, Mirian Del Carmen Hipolito Garcia, Ankur Mallick, Daniel Madrigal, Menglin Xia, Victor Rühle, Qingyun Wu, Chi Wang

    ICLR Workshop · 2025

  6. 2025

    OCCAM: Towards Cost-Efficient and Accuracy-Aware Classification Inference

    Dujian Ding, Bicheng Xu, Laks V.S. Lakshmanan

    ICLR · 2025

  7. 2024

    Pruning Attention Heads with Almost-sure Sparsity Targets

    Dujian Ding, Ganesh Jawahar, Laks V.S. Lakshmanan

    TMLR · 2024

  8. 2024

    LLM Performance Predictors are good initializers for Architecture Search

    Ganesh Jawahar, Muhammad Abdul-Mageed, Laks V.S. Lakshmanan, Dujian Ding

    Findings of ACL · 2024

  9. 2024

    Hybrid LLM: Cost-Efficient and Quality-Aware Query Routing

    Dujian Ding, Ankur Mallick, Chi Wang, Robert Sim, Subhabrata Mukherjee, Victor Rühle, Laks V.S. Lakshmanan, Ahmed Hassan Awadallah

    ICLR · 2024

  10. 2023

    On Efficient Approximate Queries over Machine Learning Models

    Dujian Ding, Sihem Amer-Yahia, Laks V.S. Lakshmanan

    VLDB · 2023

    Paper Code
  11. 2019
  12. 2019

    Collaborative Intra-tumor Heterogeneity Detection

    Sahand Khakabimamaghani, Salem Malikic, Jeffrey Tang, Dujian Ding, Ryan Morin, Leonid Chindelevitch, Martin Ester

    Bioinformatics · 2019

    Paper Code