PeopleofSuperintelligence

Researcher

Li-wei H. Lehman

Also known as Li-wei Lehman

Lehman is a research scientist at MIT's Institute for Medical Engineering and Science who develops machine learning methods for intensive care, from vital-sign dynamics to treatment-outcome prediction. With a master's degree from the Georgia Institute of Technology and a PhD from MIT, they published a switching-dynamics model of heart rate and blood pressure in 2015, and the G-Net approach to counterfactual treatment prediction in 2021. Their papers won AMIA Distinguished Paper Awards in 2020 and 2024. In 2023, they were among the authors of the MIMIC-IV database paper and first author of the VTaC alarm benchmark at NeurIPS. They lead a project on learning optimal dynamic treatment strategies from observational health data.

Timeline

  1. 2012

    Shared dynamics in ICU signals

    They hold a master's degree in computer science from the Georgia Institute of Technology and a PhD from MIT. In 2012, they publish "Discovering shared dynamics in physiological signals: Application to patient monitoring in ICU" with Shamim Nemati, Ryan P. Adams and Roger Mark at the IEEE Engineering in Medicine and Biology Society conference in San Diego.

  2. 2015

    Vital-sign dynamics and ICU mortality

    A paper with Ryan P. Adams, Louis Mayaud, George B. Moody, Atul Malhotra, Roger G. Mark and Shamim Nemati appears in the IEEE Journal of Biomedical and Health Informatics in May 2015. It uses a switching vector autoregressive framework to learn dynamics of heart rate and blood pressure; in an ICU cohort of over 450 patients from the MIMIC II database, the discovered dynamics are significantly associated with hospital mortality.

  3. 2020

    An AMIA distinguished paper award

    In 2020, the paper "Is Deep Reinforcement Learning Ready for Practical Applications in Healthcare? A Sensitivity Analysis of Duel-DDQN for Sepsis Treatment," written with MingYu Lu, Zachary Shahn, Daby Sow and Finale Doshi-Velez, wins the Distinguished Paper Award at the AMIA Annual Symposium.

  4. 2021–2022

    G-Net for counterfactual prediction

    In 2021, Lehman and colleagues publish G-Net, a recurrent network approach to g-computation for counterfactual prediction under a dynamic treatment regime, in Machine Learning for Health. In February 2022, Lehman, an MIT-IBM Watson AI Lab project lead, calls G-Net the first deep-learning approach based on g-computation that can predict both population-level and individual-level treatment effects under dynamic and time-varying treatment strategies.

  5. 2023

    MIMIC-IV and the VTaC benchmark

    In 2023, MIMIC-IV, a publicly available database sourced from the electronic health record of the Beth Israel Deaconess Medical Center, appears in Scientific Data with Lehman among its MIT authors. At NeurIPS 2023, Lehman is first author of VTaC, an annotated benchmark of over 5,000 waveform recordings of ventricular tachycardia alarms triggered by ICU bedside monitors.

  6. 2024

    A second AMIA paper award

    In March 2024, the paper "Counterfactual Sepsis Outcome Prediction under Dynamic and Time-Varying Treatment Regimes" receives the Clinical Research Informatics Distinguished Paper Award at the AMIA 2024 Informatics Summit. In August 2024, they present G-Transformer, a model for counterfactual outcome prediction under dynamic and time-varying treatment regimes, at the Machine Learning for Healthcare conference in Toronto.

  7. As of October 2026

    Machine learning for health at IMES

    Lehman is a Research Scientist at MIT's Institute for Medical Engineering and Science and a Principal Investigator leading the project Learning Optimal Dynamic Treatment Strategies from Observational Health Data. In May 2026, the paper SL-S4Wave, on self-supervised learning of physiological waveforms with structured state space models, is accepted at TMLR, and the group's postdoc opening is supported by a four-year NIH grant for 2025–2029.

Career

  1. NowMassachusetts Institute of TechnologyResearch Scientist
  2. IBM ResearchProject Lead

Education

  1. Georgia Institute of TechnologyMaster's degree
  2. Massachusetts Institute of TechnologyPhD

From the biography.

  1. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Massachusetts Institute of Technology. We estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.
  2. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Massachusetts Institute of Technology. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  3. Google had the most people shaping AI every year from 2014 to 2025Includes Massachusetts Institute of Technology. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  4. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Massachusetts Institute of Technology. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.

Sources and credits

Photo: We have no published photo of Li-wei H. Lehman yet. Send us a free-licensed photo.

  1. Li-wei H. Lehman web.mit.edu. Checked 21 October 2026.
  2. A Physiological Time Series Dynamics-Based Approach to Patient Monitoring and Outcome Prediction pmc.ncbi.nlm.nih.gov. Checked 21 October 2026.
  3. Deep-learning technique predicts clinical treatment outcomes news.mit.edu. Checked 21 October 2026.
  4. MIMIC-IV, a freely accessible electronic health record dataset pmc.ncbi.nlm.nih.gov. Checked 21 October 2026.
  5. VTaC: A Benchmark Dataset of Ventricular Tachycardia Alarms from ICU Monitors proceedings.neurips.cc. Checked 21 October 2026.
  6. MIT Research Positions in Machine Learning for Health web.mit.edu. Checked 21 October 2026.