PeopleofSuperintelligence

Researcher

Faisal Ladhak

Faisal Ladhak is an AI researcher whose work connects natural-language generation and evaluation with spatial reasoning. In 2013, he coauthored a deep-learning paper on cancer diagnosis; by July 2020, he was completing a computer science PhD at Columbia after nearly six years as an Amazon senior applied scientist. He coauthored WikiLingua in 2020, GEM in 2021, and HELM in 2022; his paper on test-set contamination earned an ICLR 2024 Honorable Mention, and he coauthored SWAN in 2025. Since 2024, he is a researcher at NVIDIA, where he builds natural-language generation systems and evaluates hallucinations and bias. In June 2026, he coauthored SpaceTools for spatial reasoning; in October 2026, he coauthored One Model to Loop Them All, which lets language models trade inference compute for accuracy.

Now at
NVIDIA

Timeline

  1. 2013

    Deep learning for cancer diagnosis

    In 2013, he coauthors Using Deep Learning to Enhance Cancer Diagnosis and Classification with Rasool Fakoor, Azade Nazi, and Manfred Huber.

  2. By July 2020

    From Amazon to Columbia

    By July 2020, he is completing a computer science PhD at Columbia after nearly six years as a senior applied scientist at Amazon. At Columbia, he works in the Computer Science Department's Natural Language Processing group under advisor Kathleen McKeown.

  3. 2020

    WikiLingua across 18 languages

    In 2020, he coauthors WikiLingua, a multilingual dataset for evaluating cross-lingual abstractive summarization. It contains article and summary pairs in 18 languages from WikiHow.

  4. 2021

    The GEM benchmark

    In 2021, he coauthors the GEM Benchmark, a living benchmark for natural language generation, evaluation, and metrics. It lets models be applied across tasks and lets researchers test evaluation strategies.

  5. 2022

    Holistic evaluation of language models

    In 2022, he coauthors Holistic Evaluation of Language Models (HELM), a benchmark designed to improve transparency about language-model capabilities, limitations, and risks.

  6. 2024

    An ICLR honorable mention

    In 2024, he coauthors Proving Test Set Contamination in Black-Box Language Models, which receives an ICLR Honorable Mention. The paper proposes a procedure that detects test-set contamination without pretraining data or model weights and gives exact false-positive guarantees.

  7. 2025

    Long-context modeling with SWAN

    In 2025, he coauthors SWAN, a causal Transformer in the decoder-only style that generalizes to sequences substantially longer than those seen during training. Its design lowers training cost and increases inference throughput.

  8. Since 2024

    Research at NVIDIA

    Since 2024, he is a researcher at NVIDIA who builds natural-language generation systems and evaluates hallucinations and bias. In June 2026, he coauthors SpaceTools, a CVPR paper with NVIDIA's Learning and Perception Research group; the project equips vision-language models with vision and robotic tools for spatial reasoning. In October 2026, he coauthors One Model to Loop Them All, a training recipe that lets language models trade inference compute for accuracy by changing the loop count of a shared layer block.

Career

  1. Since 2024NVIDIAResearcher
  2. Until 2020AmazonSenior applied scientist

Education

  1. Columbia UniversityPhD

From the biography.

  1. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Faisal Ladhak. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  2. Google has the highest combined h-index among AI organizationsIncludes NVIDIA. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  3. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes NVIDIA. 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.
  4. Universities (31%) and startups and other companies (25%) employ the largest shares of 2020–2026 entrantsIncludes NVIDIA. Among 819 people whose first recorded role started in 2020–2026 and who have a known current organization, 60% work in commercial sectors (25% at startups and other companies, 23% at frontier labs and 13% at big tech, with 9 in two), and 31% work at universities, compared with 66% for earlier entrants.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

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

  1. Faisal Ladhak cs.columbia.edu. Checked 22 October 2026.
  2. Amazon at ACL: Old standards and new forays amazon.science. Checked 22 October 2026.
  3. WikiLingua: A New Benchmark Dataset for Cross-Lingual Abstractive Summarization aclanthology.org. Checked 22 October 2026.
  4. The GEM Benchmark: Natural Language Generation, its Evaluation and Metrics aclanthology.org. Checked 22 October 2026.
  5. Holistic Evaluation of Language Models arxiv.org. Checked 22 October 2026.
  6. ICLR Oral Proving Test Set Contamination in Black-Box Language Models iclr.cc. Checked 22 October 2026.
  7. SWAN: An Efficient and Scalable Approach for Long-Context Language Modeling aclanthology.org. Checked 22 October 2026.
  8. Faisal Ladhak — Lacuna lacuna.tiptreesystems.com. Checked 22 October 2026.
  9. SpaceTools: Tool-Augmented Spatial Reasoning via Double Interactive RL | NVIDIA Learning and Perception Research research.nvidia.com. Checked 22 October 2026.
  10. SpaceTools: Tool-Augmented Spatial Reasoning via Double Interactive RL spacetools.github.io. Checked 22 October 2026.
  11. One Model to Loop Them All: Elastic Latent Recurrence for Language Models arxiv.org. Checked 22 October 2026.