PeopleofSuperintelligence

Researcher

Ajay Patel

Several of Ajay Patel's papers use synthetic data to train language and vision-language models. He co-founded Plasticity in 2016, a company that built natural language processing APIs for developers and was acquired in October 2020. He co-wrote a paper on prompting bidirectional models that ICLR presented in 2023, and DataDreamer, a tool for synthetic data generation and reproducible LLM workflows, which ACL published in 2024. He co-authored Molmo and PixMo, open vision-language models posted to arXiv in September 2024, and CoSyn, which ACL published in 2025. In 2026 he co-authored FineInstructions, a paper on scaling synthetic instructions to pre-training scale.

Timeline

  1. 2016–2020

    Co-founding Plasticity

    He co-founds Plasticity in 2016, a company that builds natural language processing APIs for developers, and it is acquired in October 2020.

  2. 2023

    SAP prompting for bidirectional models

    Bidirectional Language Models Are Also Few-shot Learners, which he co-writes, presents SAP prompting for bidirectional models at ICLR in 2023.

  3. 2024

    DataDreamer for LLM workflows

    In 2024 ACL publishes DataDreamer, a tool he co-writes for synthetic data generation and reproducible LLM workflows.

  4. September 2024

    Molmo and PixMo open models

    The Molmo and PixMo paper, which he co-authors, is posted to arXiv in September 2024.

  5. 2025

    CoSyn for synthetic text-rich data

    CoSyn, a paper he co-authors, uses code from text-only language models to create synthetic text-rich image data and appears at ACL 2025.

  6. 2025

    Style embeddings from synthetic examples

    StyleDistance, which he co-writes, trains style embeddings on synthetic parallel examples, and the NAACL 2025 proceedings publish it.

  7. 2026

    Scaling synthetic instructions

    He co-authors FineInstructions, a paper on scaling synthetic instructions to pre-training scale, which is posted to arXiv in January 2026 and listed for COLM 2026.

Sources and credits

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

  1. Plasticity: Natural language processing APIs for developers. | Y Combinator ycombinator.com. Checked 14 October 2026.
  2. About | Ajay Patel ajayp.app. Checked 14 October 2026.
  3. Ajay Patel ajayp.app. Checked 14 October 2026.
  4. ICLR Poster Bidirectional Language Models Are Also Few-shot Learners iclr.cc. Checked 14 October 2026.
  5. DataDreamer: A Tool for Synthetic Data Generation and Reproducible LLM Workflows - ACL Anthology aclanthology.org. Checked 14 October 2026.
  6. [2409.17146] Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models arxiv.org. Checked 14 October 2026.
  7. Chris Callison-Burch cis.upenn.edu. Checked 14 October 2026.
  8. [2502.14846] Scaling Text-Rich Image Understanding via Code-Guided Synthetic Multimodal Data Generation arxiv.org. Checked 14 October 2026.
  9. StyleDistance: Stronger Content-Independent Style Embeddings with Synthetic Parallel Examples - ACL Anthology aclanthology.org. Checked 14 October 2026.
  10. [2601.22146] FineInstructions: Scaling Synthetic Instructions to Pre-Training Scale arxiv.org. Checked 14 October 2026.