Luca Soldaini

Career

Luca Soldaini completed a PhD in computer science at Georgetown University in 2018, in the Information Retrieval Lab with Nazli Goharian, with a thesis on information retrieval systems for medical experts and lay health users. Soldaini then worked as a senior applied scientist at Alexa on efficient question answering for long-tail knowledge, including work in 2021 to scale generative question answering in production.¹

From 2022 to early 2026 Soldaini was a lead research scientist at Ai2, co-leading the Olmo project. Soldaini is now a member of the technical staff at Microsoft AI, working on the MAI-Thinking models.¹²

Open models and data

Soldaini is first author of Dolma (2024), a three-trillion-token open corpus for language model pretraining research, and a co-author of the first OLMo paper, "OLMo: Accelerating the Science of Language Models" (2024). During Soldaini's time at Ai2, the Olmo project released three generations of dense, mixture-of-experts, hybrid and multimodal models with their data, code, recipes and checkpoints.³⁴¹

Soldaini is second-listed author of OLMoE (2024), an open mixture-of-experts language model, and of "2 OLMo 2 Furious" (2024), the OLMo 2 report, and co-authored olmOCR (2025), which uses vision language models to extract text from PDF documents.⁵⁶⁷

Work and contributions

• OLMo

  Co-led the Olmo project at Ai2, which released dense, mixture-of-experts, hybrid, and multimodal open models alongside their data, code, recipes, and checkpoints.¹

• Dolma

  First author of the paper describing Dolma, a three-trillion-token open corpus built for language model pretraining research, released with an open-source data curation toolkit.³

• Olmo 3 7B (base) Open weights

  Listed in the Olmo 3 report's author contributions for technical leadership, model architecture and training methodology, and base model data acquisition.⁸

Timeline

1. 2018Completed a PhD in computer science at Georgetown University's Information Retrieval Lab.¹
2. 2021Helped scale generative question answering at Alexa as a senior applied scientist.¹
3. 2022Joined Ai2 as a lead research scientist; co-led the Olmo project until early 2026.¹
4. Jan 31, 2024First author of Dolma, an open corpus of three trillion tokens for pretraining research.³
5. Feb 1, 2024Co-authored "OLMo: Accelerating the Science of Language Models".⁴
6. Sep 3, 2024Second-listed author of OLMoE, an open mixture-of-experts language model.⁵
7. Dec 31, 2024Second-listed author of "2 OLMo 2 Furious", the OLMo 2 technical report.⁶
8. Feb 25, 2025Co-authored olmOCR, on extracting text from PDFs with vision language models.⁷

Other documented roles

Roles the sources document that are not confirmed as current.

• Lead research scientist, Ai2 (Allen Institute for AI)¹

Where they work

• Microsoft 75 people in the wiki

Related people

• Ronen EldanAlso at Microsoft
• Rangan MajumderAlso at Microsoft
• Yuanzhi LiAlso at Microsoft
• Bill DolanAlso at Microsoft
• Kristina ToutanovaAlso at Microsoft
• Leo GaoLanguage models · Datasets and data
• Jesse DodgeDatasets and data · Language models
• Sridhar RamaswamyLanguage models · Datasets and data
• Avi SilLanguage models · Datasets and data
• Christopher RéLanguage models · Datasets and data