PeopleofSuperintelligence

Researcher

Germain Kolossov

Germain Kolossov is a machine learning researcher whose work studies training-data selection and diffusion-based generation. In 2023, he coauthored a paper with Andrea Montanari and Pulkit Tandon on data selection under weak supervision. The work was presented orally at ICLR and received an Honorable Mention in 2024; their experiments found cases where training on 40% of a dataset had a lower error rate than training on the full dataset. In 2026, he coauthored research on generating random solutions to k-SAT and k-XORSAT formulas. By September 2026, he is a researcher affiliated with Granica Computing Inc.

Timeline

  1. 2023

    A theory of data selection

    In 2023, he coauthors “Towards a statistical theory of data selection under weak supervision” with Andrea Montanari and Pulkit Tandon. The paper selects a smaller sample, acquires its labels, and uses them to train a model.

  2. 2024

    An oral ICLR presentation

    In 2024, his data-selection paper is presented orally at ICLR.

  3. 2024

    Training on less data

    In 2024, he and his coauthors test their selection method on a real autonomous-vehicle dataset; in some cases, training on 40% of the original data yields a lower error rate than training on the full set.

  4. 2024

    An ICLR honorable mention

    ICLR gives an Honorable Mention in 2024 to the data-selection paper he coauthors with Andrea Montanari and Pulkit Tandon.

  5. 2026

    Diffusion for discrete generation

    In 2026, he coauthors “Generating from Discrete Distributions Using Diffusions: Insights from Random Constraint Satisfaction Problems.” The paper studies generating uniformly random solutions to random k-SAT and k-XORSAT formulas.

  6. By September 2026

    Research at Granica

    By September 2026, he is a researcher affiliated with Granica Computing Inc.

  1. Google had the most people shaping AI every year from 2014 to 2025Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  2. The United States employs an estimated 48% of people shaping AIWe estimate where the people shaping AI work today, how universities, companies and research institutes split across countries, and how country shares have changed since the 1960s.
  3. 43% of living people shaping AI work outside their country of birthWe estimate that 43% of living people shaping AI work outside their country of birth (520 of 1,198 people with known birth and employment countries).
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionWe 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 Germain Kolossov yet. Send us a free-licensed photo.

  1. [2309.14563] Towards a statistical theory of data selection under weak supervision arxiv.org. Checked 21 October 2026.
  2. Improving AI via optimal selection of training samples granica.ai. Checked 21 October 2026.
  3. ICLR 2024 Awards iclr.cc. Checked 21 October 2026.
  4. [2603.20589] Generating from Discrete Distributions Using Diffusions: Insights from Random Constraint Satisfaction Problems arxiv.org. Checked 21 October 2026.
  5. Generating from Discrete Distributions Using Diffusions: Insights from Random Constraint Satisfaction Problems ar5iv.labs.arxiv.org. Checked 21 October 2026.