PeopleofSuperintelligence

Builder

Paula Kurylowicz

Paula Kurylowicz is an AI model contributor whose work runs from Google's Gemini report to Mistral AI's reasoning and speech models. In 2023 she co-authored a paper on mallocMuZero, a reinforcement-learning agent that finds faster memory mappings for machine-learning compilers, and was among the authors of the Gemini report. A 2024 Google DeepMind paper on reinforcement learning for tokamak magnetic control also has her as a co-author. In 2025 she was among the authors of Magistral, Mistral's first reasoning model, and of the Voxtral audio chat models. She is an author of the 2026 papers on Ministral 3, Voxtral Realtime, Voxtral TTS, Robostral Navigate and Shieldstral.

Works
11

Publication numbers from OpenAlex, as of 22 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. May 2023

    Memory mapping with mallocMuZero

    In May 2023, she co-authors a paper, with Oriol Vinyals and Daniel J. Mankowitz among others, that frames memory mapping in machine-learning compilers as a single-player game called mallocGame. Its reinforcement-learning agent, mallocMuZero, discovers memory mappings that lead to faster execution times on real machine-learning workloads.

  2. December 2023

    The Gemini report

    In December 2023, she is among the authors of the Gemini report from the Gemini Team at Google. The report introduces the Gemini family of multimodal models in Ultra, Pro and Nano sizes.

  3. March 2024

    Tokamak magnetic control

    In March 2024, she co-authors a Google DeepMind paper in Fusion Engineering and Design on reinforcement learning for plasma magnetic control. It builds on Degrave et al. (2022), shows up to 65% better shape accuracy in simulation, and tests the upgraded controllers on the TCV tokamak.

  4. June 2025

    Mistral's first reasoning model

    In June 2025, she is among the authors of Magistral, Mistral's first reasoning model, built with the company's own scalable reinforcement-learning pipeline. Magistral Medium is trained on top of Mistral Medium 3 with reinforcement learning alone, and Magistral Small is open-sourced under Apache 2.0.

  5. July 2025

    Voxtral audio chat models

    In July 2025, she is among the authors of Voxtral Mini and Voxtral Small, two multimodal audio chat models released under the Apache 2.0 license. A 32K context window lets them handle audio files up to 40 minutes long.

  6. January 2026

    The Ministral 3 family

    In January 2026, she is among the authors of Ministral 3, a family of dense language models in 3B, 8B and 14B sizes with image understanding, released under Apache 2.0. The models are derived through Cascade Distillation, an iterative pruning and continued training technique.

  7. February–March 2026

    Voxtral Realtime and Voxtral TTS

    In February 2026, she is among the Mistral-AI authors of Voxtral Realtime, a natively streaming speech-recognition model that performs on par with Whisper at a delay of 480ms. In March 2026, she is among the authors of Voxtral TTS, a multilingual text-to-speech model that generates speech from as little as 3 seconds of reference audio.

  8. July 2026

    Robostral Navigate and Shieldstral

    In July 2026, she is among the authors of Robostral Navigate, an 8B vision-language model that navigates robots using only a stream of monocular RGB images. She is also an author of Shieldstral, a 3B multimodal safety classifier.

  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 Paula Kurylowicz yet. Send us a free-licensed photo.

  1. Optimizing Memory Mapping Using Deep Reinforcement Learning arxiv.org. Checked 20 October 2026.
  2. Gemini: A Family of Highly Capable Multimodal Models arxiv.org. Checked 20 October 2026.
  3. Towards Practical Reinforcement Learning for Tokamak Magnetic Control — Google DeepMind deepmind.google. Checked 20 October 2026.
  4. Magistral arxiv.org. Checked 20 October 2026.
  5. Magistral | ArxivLens arxivlens.com. Checked 20 October 2026.
  6. Voxtral arxiv.org. Checked 20 October 2026.
  7. Ministral 3 arxiv.org. Checked 20 October 2026.
  8. Voxtral Realtime arxiv.org. Checked 20 October 2026.
  9. Voxtral TTS arxiv.org. Checked 20 October 2026.
  10. Robostral Navigate arxiv.org. Checked 20 October 2026.
  11. Shieldstral arxiv.org. Checked 20 October 2026.