PeopleofSuperintelligence

Researcher

Saurabh Dash

Saurabh Dash is a researcher at Cohere Labs whose research covers multimodality, reinforcement learning and large-language-model efficiency. His education includes undergraduate study at IIT Kharagpur and PhD study at Georgia Tech with advisor Saibal Mukhopadhyay; he interned on Apple’s MIND team in 2021 and coauthored ICLR papers on spiking neural networks and event-based perception. He coauthored research on large-model quantization in 2023 and multilingual quantization in 2024, then coauthored Aya Vision and the Command A report in 2025. In 2026, he coauthored Tiny Aya, Soft-SVeRL and CALIBER, and he mentored a Tiny Aya Expedition team studying language-aware quantization.

Now at
Cohere
Works
10
Citations
35
h-index
3

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 2021–2023

    Apple internship and ICLR papers

    His academic background includes undergraduate study at IIT Kharagpur and PhD study at Georgia Tech with advisor Saibal Mukhopadhyay. In summer 2021, he works as a machine-learning research intern on Apple’s MIND team, supervised by Hema Koppula. He coauthors ICLR papers on spiking neural networks in 2022 and event-based perception in 2023.

  2. 2023

    Quantization at scale

    In 2023, he coauthors Intriguing Properties of Quantization at Scale. The study finds that outlier dimensions are sensitive to pretraining optimization conditions and reports quantized models from 410M to 52B parameters with minimal performance loss.

  3. 2024

    Multilingual model quantization

    In 2024, he coauthors How Does Quantization Affect Multilingual LLMs? The study finds that languages using non-Latin scripts are most affected and mathematical reasoning degrades fastest.

  4. 2025

    Aya Vision and Command A

    In 2025, he coauthors Aya Vision: Advancing the Frontier of Multilingual Multimodality, a vision-language family in 8B and 32B sizes. A 2025 report lists him as an author of Command A and describes it as an agent-optimized multilingual model for enterprise use across 23 business languages.

  5. 2026

    Tiny Aya across 70 languages

    In 2026, he coauthors Tiny Aya: Bridging Scale and Multilingual Depth. The 3.35B-parameter model is trained on 70 languages for translation, multilingual understanding and target-language generation.

  6. 2026

    Soft-SVeRL and self-verification

    In May 2026, he coauthors Soft-SVeRL, a self-verifying reinforcement-learning method in which the policy also acts as its verifier. The paper studies language-model training with decomposed verification signals and partial-credit rewards.

  7. 2026

    CALIBER confidence calibration

    In June 2026, he coauthors CALIBER, a method that trains reasoning models to estimate confidence before and after an answer. The study matches each confidence estimate to its own prediction target.

  8. As of October 2026

    Researcher at Cohere Labs

    As of October 2026, he is a Researcher at Cohere Labs, where he works on multimodality, reinforcement learning and language-model efficiency. In July 2026, he mentors a Tiny Aya Expedition team studying language-aware quantization.

Career

  1. Since 2024CohereResearcher
  2. 2021AppleMachine-learning research intern

Education

  1. Indian Institute of Technology KharagpurUndergraduate study
  2. Georgia Institute of TechnologyPhD

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. How Does Quantization Affect Multilingual LLMs?2024 · 14 citationsWith Sara Hooker and Sebastian Ruder
  2. Intriguing Properties of Quantization at Scale2023 · neural information processing systems · 11 citationsWith Arash Ahmadian, Phil Blunsom and Sara Hooker
  3. Aya 23: Open Weight Releases to Further Multilingual Progress2024 · 9 citationsWith Phil Blunsom, Sara Hooker and Sebastian Ruder
  4. Aya Vision: Advancing the Frontier of Multilingual Multimodality2025 · 1 citationsWith Arash Ahmadian, Beyza Ermis, Phil Blunsom, Sara Hooker and Shivalika Singh
  5. CALIBER: Calibrating Confidence Before and After Reasoning in Language Models2026 · 0 citationsWith Beyza Ermis

From OpenAlex, as of 21 October 2026.

  1. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Saurabh Dash. 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. Among employers with 30+ full-time cases, OpenAI leads primary technical role H-1B pay at $310,000, and Amazon, Google and Microsoft file 57% of all casesIncludes Cohere. Across 440,669 certified US H-1B filings from 44 employers (FY2020–FY2026), OpenAI leads primary technical role base pay at $310,000 and xAI leads overall base pay at $290,000.
  3. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Cohere. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  4. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Cohere. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.

Colleagues

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

Sources and credits

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

  1. Saurabh Dash | GT saurabhdash.com. Checked 21 October 2026.
  2. Intriguing Properties of Quantization at Scale arxiv.org. Checked 21 October 2026.
  3. How Does Quantization Affect Multilingual LLMs? - ACL Anthology aclanthology.org. Checked 21 October 2026.
  4. A Deepdive into Aya Vision: Advancing the Frontier of Multilingual Multimodality huggingface.co. Checked 21 October 2026.
  5. Aya Vision: Advancing the Frontier of Multilingual Multimodality arxiv.org. Checked 21 October 2026.
  6. Command A: An Enterprise-Ready Large Language Model arxiv.org. Checked 21 October 2026.
  7. Tiny Aya: Bridging Scale and Multilingual Depth arxiv.org. Checked 21 October 2026.
  8. Soft-SVeRL: Self-Verified Reinforcement Learning with Soft Rewards arxiv.org. Checked 21 October 2026.
  9. CALIBER: Calibrating Confidence Before and After Reasoning in Language Models arxiv.org. Checked 21 October 2026.
  10. Tiny Aya in the wild: What people are building with open multilingual AI cohere.com. Checked 21 October 2026.