PeopleofSuperintelligence

Researcher

Jiaming Song

Jiaming Song is a generative AI researcher whose work spans diffusion sampling and multimodal foundation models. He completed undergraduate study at Tsinghua in 2016, coauthored DDIM, and earned a Stanford PhD under Stefano Ermon. His paper Comparing Distributions by Measuring Differences that Affect Decision Making received an ICLR 2022 Outstanding Paper Award, and he coauthored Luma's March 2025 article introducing Inductive Moment Matching for more efficient sampling. By September 2024, he was Luma AI's Chief Scientist after research at NVIDIA and a Stanford postdoc; he presented Dream Machine at Stanford. As of October 2026, he builds multimodal AI systems for general intelligence, working on models that reason across modalities, generate rich worlds, and turn human intent into controllable outputs.

Timeline

  1. 2016

    Tsinghua undergraduate

    He completes undergraduate study at Tsinghua in 2016.

  2. 2020–2021

    DDIM speeds diffusion sampling

    In 2020, he coauthors Denoising Diffusion Implicit Models with Chenlin Meng and Stefano Ermon; the paper appears at ICLR 2021. It introduces non-Markovian diffusion processes whose reverse process can be much faster to sample.

  3. 2022

    An ICLR Outstanding Paper

    In 2022, his paper Comparing Distributions by Measuring Differences that Affect Decision Making receives an ICLR Outstanding Paper Award.

  4. September 2024

    Chief Scientist at Luma

    By September 2024, he holds a Stanford PhD under Stefano Ermon. Before Luma, he works as a Stanford postdoc and research scientist at NVIDIA; at Luma, he is Chief Scientist, working on next-generation multimodal foundation models.

  5. September 2024

    Dream Machine at Stanford

    At a Stanford event in September 2024, he presents Dream Machine, a Luma AI video foundation model that explores visual capabilities such as depth, segmentation, light transport, dynamics and causality.

  6. March 2025

    Inductive Moment Matching

    In March 2025, he coauthors Luma's article introducing Inductive Moment Matching (IMM), a pre-training method that reports over a tenfold increase in sampling efficiency compared with diffusion models.

  7. By 2026

    Genie, Ray and Uni-1

    By 2026, his Luma research spans shifts from 3D generation models with Genie to video generation with Ray and unified multimodal models with Uni-1; it covers model architecture, training infrastructure, and data processing.

  8. As of October 2026

    Building multimodal AI

    As of October 2026, he builds multimodal AI systems for general intelligence; his work focuses on foundation models that reason across modalities, generate rich worlds, and turn human intent into controllable outputs.

Career

  1. Stanford UniversityPostdoc
  2. NVIDIAResearch Scientist
  3. Luma AIChief Scientist

Education

  1. 2016Tsinghua UniversityUndergraduate study
  2. Stanford UniversityPhD

From the biography.

Videos

YouTube videos in which Jiaming Song speaks or appears, newest first. A model chose them from YouTube search results. List updated 21 October 2026.

  1. Luma Labs' Diffusion Revolution: from Dream Machine to Multimodal Worldsim - Amit Jain, Jiaming SongCognitive Revolution "How AI Changes Everything" · 11 May 2025 · 1 h 19 min

Video data from YouTube API Services. By watching you agree to the YouTube Terms of Service. How we use YouTube data.

  1. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Jiaming Song. 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.
  2. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Stanford University. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.
  3. Google had the most people shaping AI every year from 2014 to 2025Includes Stanford University. 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. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Stanford University. 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.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Jiaming Song yet. Send us a free-licensed photo.

  1. Jun Zhu's Homepage ml.cs.tsinghua.edu.cn. Checked 21 October 2026.
  2. [2010.02502] Denoising Diffusion Implicit Models arxiv.org. Checked 21 October 2026.
  3. ICLR 2022 Awards iclr.cc. Checked 21 October 2026.
  4. Dream Machine: Emergent Capabilities from Video Foundation Models ee.stanford.edu. Checked 21 October 2026.
  5. Jiaming Song tsong.me. Checked 21 October 2026.
  6. Breaking the Algorithmic Ceiling in Pre-Training with Inductive Moment Matching lumalabs.ai. Checked 21 October 2026.