PeopleofSuperintelligence

Researcher

Daniel D'souza

Daniel D'souza is a Research Engineer at Cohere whose work spans multilingual AI, machine-learning fairness, and coding-agent teams. He studied electrical and communications engineering at KLS Gogte Institute of Technology, pursued graduate study at the University of Michigan, and worked at ProQuest. His research includes African-language entity recognition, subgroup performance in neural-network ensembles, the Aya model family, and multilingual inference-time scaling. As of October 2026, he studies how coding-agent teams can solve software tasks.

Now at
Cohere
Works
26
Citations
567
h-index
8
Badges

Publication numbers from OpenAlex, as of 19 October 2026.

Timeline

  1. 2009–2018

    Engineering and graduate study

    He earns a Bachelor of Engineering in Electrical, Electronics and Communications Engineering at KLS Gogte Institute of Technology in 2013. At the University of Michigan, he pursues an MS in 2016–2017 and works as a research assistant in the Language and Information Technologies Lab in 2018.

  2. 2017–2024

    ProQuest data science

    At ProQuest, he is a software engineering intern in 2017 and a data scientist in 2018. He serves as a senior data scientist from 2022 to 2024.

  3. 2021–2023

    African-language NER and fairness

    He coauthors MasakhaNER, a dataset for named entity recognition in ten African languages, in 2021. In 2023, he coauthors FAIR-Ensemble, which finds that homogeneous deep-neural-network ensembles can improve minority-group performance as the number of models grows.

  4. 2024

    Building multilingual Aya models

    He coauthors Aya Model, a massively multilingual generative language model that follows instructions in 101 languages, and Aya Dataset, a human-curated instruction collection spanning 65 languages. He also coauthors Aya Expanse, a family of 8B and 32B parameter multilingual language models.

  5. 2025

    Scaling inference across languages

    His 2025 paper studies inference-time compute for open 8B and 111B models across seven languages. The team proposes language- and task-aware sampling strategies that improve performance across languages and tasks.

  6. As of October 2026

    Research Engineer at Cohere

    As of October 2026, he is a Research Engineer at Cohere. With coauthors Keenan Samway and Nouha Dziri, he studies teams of coding agents. On ProgramBench, the team finds that assigning workers to exploration and planning matches the performance of delegating planning, implementation, and verification together at half the cost.

Career

  1. Since 2024CohereResearch Engineer
  2. 2018University of MichiganResearch Assistant
  3. 2017–2024ProQuestSoftware Engineering Intern, Data Scientist

Education

  1. 2017University of MichiganMS
  2. 2013KLS Gogte Institute of TechnologyBachelor of Engineering in Electrical, Electronics and Communications Engineering

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

Most-cited works

  1. MasakhaNER: Named Entity Recognition for African Languages2021 · Transactions of the Association for Computational Linguistics · 258 citationsWith Sebastian Ruder
  2. Estimating Example Difficulty using Variance of Gradients2022 · IEEE Computer Society Conference on Computer Vision and Pattern Recognition · 68 citationsWith Sara Hooker
  3. sourmash v4: A multitool to quickly search, compare, and analyze genomic and metagenomic data sets2024 · The Journal of Open Source Software · 66 citations
  4. Aya Model: An Instruction Finetuned Open-Access Multilingual Language Model2024 · Annual Meeting of the Association for Computational Linguistics (ACL) · 52 citationsWith Niklas Muennighoff, Phil Blunsom and Sara Hooker
  5. Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning2024 · Annual Meeting of the Association for Computational Linguistics (ACL) · 29 citationsWith Niklas Muennighoff, Sara Hooker and Sebastian Ruder

From OpenAlex, as of 19 October 2026.

  1. Google alumni founded or led 87 other organizations, the most among the labs we compareIncludes Cohere. We trace 212 organizations founded or led by AI lab alumni, and 223 career moves between the labs.
  2. The company share of AI researchers’ top papers rose from 2% to 32%Includes Cohere. We estimate where leading AI researchers worked when their most-cited papers appeared. The company share rose from 2% in 1975–1979 to 32% in 2020–2024.
  3. Google has the highest combined h-index among AI organizationsIncludes Cohere. Our estimates of combined and median h-index at AI organizations, using current affiliations and OpenAlex records.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes University of Michigan. Our estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.

Colleagues

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

Sources and credits

Photo: We have no published photo of Daniel D'souza yet. Send us a free-licensed photo.

  1. Daniel D'souza — Lacuna lacuna.tiptreesystems.com. Checked 19 October 2026.
  2. MasakhaNER: Named Entity Recognition for African Languages — ACL Anthology aclanthology.org. Checked 19 October 2026.
  3. FAIR-Ensemble: When Fairness Naturally Emerges From Deep Ensembling arxiv.org. Checked 19 October 2026.
  4. Aya Model: An Instruction Finetuned Open-Access Multilingual Language Model — ACL Anthology aclanthology.org. Checked 19 October 2026.
  5. Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning — ACL Anthology aclanthology.org. Checked 19 October 2026.
  6. Aya Expanse: Combining Research Breakthroughs for a New Multilingual Frontier arxiv.org. Checked 19 October 2026.
  7. When Life Gives You Samples: The Benefits of Scaling up Inference Compute for Multilingual LLMs — ACL Anthology aclanthology.org. Checked 19 October 2026.
  8. Lessons from scaling teams of coding agents: reach the frontier at 30% of the cost cohere.com. Checked 19 October 2026.