PeopleofSuperintelligence

Researcher

Aleksandra Piktus

Aleksandra Piktus is an AI researcher whose work connects language models with retrieval, training data and evaluation. She coauthored misspelling-resistant embeddings, retrieval-augmented generation and KILT, then coauthored tools that let researchers search language-model training corpora. Her work on data-constrained scaling earned an Outstanding Paper runner-up award at NeurIPS 2023, and a later study evaluated panels of language models as judges. She is a research engineer at Cohere and holds a Sapienza PhD in data science on unstructured data for large language models.

Now at
Cohere

Timeline

  1. 2019

    Misspelling-oblivious embeddings

    In 2019, she coauthors Misspelling Oblivious Word Embeddings, which combines FastText with subwords and supervised misspelling patterns to place misspelled words near their correct forms.

  2. 2020–2021

    RAG and the KILT benchmark

    In 2020, she coauthors Retrieval-Augmented Generation (RAG), which combines pre-trained parametric memory with non-parametric memory for language generation. Her KILT paper is accepted at NAACL in 2021 and grounds knowledge-intensive language tasks in a shared Wikipedia snapshot.

  3. June 2022

    Research engineer at Hugging Face

    In June 2022, she is a research engineer at Hugging Face and gives a Sapienza lecture on knowledge-intensive NLP, retrieval augmentation and RAG. She joins Hugging Face's Science team after research at Facebook AI Research in London. Earlier, she starts her career as a Facebook software engineer working on Facebook Search and misinformation.

  4. 2023

    NeurIPS Outstanding Paper runner-up

    At NeurIPS 2023, her paper Scaling Data-Constrained Language Models earns an Outstanding Paper runner-up award. Its experiments vary data repetition and compute budgets for language models.

  5. 2023

    Search tools for training data

    In 2023, she coauthors the ROOTS Search Tool, which offers fuzzy and exact search over the 1.6-terabyte multilingual ROOTS corpus used to train BLOOM. She also coauthors GAIA Search, which joins Hugging Face and Pyserini to help researchers explore large text collections before training NLP models.

  6. 2024

    Replacing judges with juries

    In 2024, she coauthors a Panel of LLM Evaluators (PoLL) that uses smaller models from different families to judge generated text. The paper reports that the panel outperforms a single large judge and costs over seven times less.

  7. 2025

    Command A language model

    In 2025, she coauthors Cohere's Command A technical report, which describes an enterprise model with multilingual support, retrieval-augmented generation and tool use.

  8. As of October 2026

    Research engineer at Cohere

    As of October 2026, she is a research engineer at Cohere working on large language models. She holds a PhD in data science from Sapienza, where her thesis examines unstructured data for large language models.

Career

  1. NowCohereResearch Engineer
  2. From 2022Hugging FaceResearch Engineer
  3. MetaSoftware Engineer

Education

  1. Sapienza University of RomePhD in data science

From the biography.

  1. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Aleksandra Piktus. 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 Aleksandra Piktus yet. Send us a free-licensed photo.

  1. Misspelling Oblivious Word Embeddings - ACL Anthology aclanthology.org. Checked 20 October 2026.
  2. [2005.11401] Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks arxiv.org. Checked 20 October 2026.
  3. [2009.02252] KILT: a Benchmark for Knowledge Intensive Language Tasks arxiv.org. Checked 20 October 2026.
  4. PhD in DATA SCIENCE - Theory and Practice of Deep Learning sites.google.com. Checked 20 October 2026.
  5. Aleksandra Piktus | Dipartimento di Ingegneria informatica, automatica e gestionale diag.uniroma1.it. Checked 20 October 2026.
  6. NeurIPS 2023 Awards neurips.cc. Checked 20 October 2026.
  7. [2302.14035] The ROOTS Search Tool: Data Transparency for LLMs arxiv.org. Checked 20 October 2026.
  8. GAIA Search: Hugging Face and Pyserini Interoperability for NLP Training Data Exploration - ACL Anthology aclanthology.org. Checked 20 October 2026.
  9. [2404.18796] Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models arxiv.org. Checked 20 October 2026.
  10. [2504.00698] Command A: An Enterprise-Ready Large Language Model arxiv.org. Checked 20 October 2026.
  11. ALEKSANDRA PIKTUS | DATA SCIENCE phd.uniroma1.it. Checked 20 October 2026.