Lab leader

Oleksii Kuchaiev

As of October 2026, Oleksii Kuchaiev is director of applied research for AI model post-training at NVIDIA. He earned a Ph.D. from the University of California, Irvine in 2010, with a dissertation titled Modeling and Alignment of Biological Networks. He co-authored Mixed Precision Training in 2017, a technique for training deep neural networks with half-precision numbers, and he was first author of the NeMo toolkit paper in 2019. Later he co-authored the Nemotron-4 340B technical report in 2024, Llama-Nemotron reasoning models in 2025, and PivotRL, a framework for agentic post-training, in 2026.

Now at
NVIDIA
Works
76

Publication numbers from OpenAlex, as of 15 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. 2010

    A dissertation on biological networks

    He earns a Ph.D. in information and computer science from the University of California, Irvine in 2010, with a dissertation titled Modeling and Alignment of Biological Networks.

  2. 2014

    A computational network toolkit

    An Interspeech 2014 invited talk introduces the computational network toolkit (CNTK), a general-purpose C++ implementation of computational networks, and he is one of its co-authors.

  3. 2017

    Training with half precision

    In 2017 he co-authors Mixed Precision Training, a technique for training deep neural networks with half-precision floating point numbers.

  4. 2019

    The NeMo toolkit

    In 2019 he is the first author of NeMo, a framework-agnostic toolkit for creating AI applications from neural modules.

  5. 2024

    The Nemotron-4 340B family

    The 2024 Nemotron-4 340B technical report, which he co-authors, releases the Nemotron-4 340B model family.

  6. 2025

    Llama-Nemotron reasoning models

    Llama-Nemotron: Efficient Reasoning Models, a 2025 paper he co-authors, introduces an open family of reasoning models.

  7. March 2026

    A framework for agentic post-training

    In 2026 he co-authors PivotRL, a framework that combines the compute efficiency of supervised fine-tuning with the out-of-domain accuracy of reinforcement learning.

  8. As of October 2026

    Director of applied research

    He is director of applied research for AI model post-training at NVIDIA.

Sources and credits

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

  1. Oleksii Kuchaiev - The Mathematics Genealogy Project mathgenealogy.org. Checked 15 October 2026.
  2. Author: Oleksii Kuchaiev | NVIDIA Technical Blog developer.nvidia.com. Checked 15 October 2026.
  3. ISCA Archive - An introduction to computational networks and the computational network toolkit (invited talk) isca-archive.org. Checked 15 October 2026.
  4. [1710.03740] Mixed Precision Training arxiv.org. Checked 15 October 2026.
  5. [1909.09577] NeMo: a toolkit for building AI applications using Neural Modules arxiv.org. Checked 15 October 2026.
  6. [2406.11704] Nemotron-4 340B Technical Report arxiv.org. Checked 15 October 2026.
  7. [2505.00949] Llama-Nemotron: Efficient Reasoning Models arxiv.org. Checked 15 October 2026.
  8. [2603.21383] PivotRL: High Accuracy Agentic Post-Training at Low Compute Cost arxiv.org. Checked 15 October 2026.
  9. NVIDIA Nemotron Summit at NeurIPS nvevents.nvidia.com. Checked 15 October 2026.