Sophie Lebrecht
Sophie Lebrecht built an AI career that connected visual neuroscience with computer vision and open vision-language models. She earned a PhD in Cognitive Science at Brown in 2012, pursued postdoctoral research at Carnegie Mellon, and co-founded Neon Labs, turning her visual-preference work into thumbnail-selection software. The World Economic Forum named Neon a Technology Pioneer in 2015, and at Xnor.ai she directed operations for computer vision while the company pursued a 60-frame-per-second image-tracking goal before Apple's early 2020 acquisition. She coauthored the open Molmo and PixMo work in 2024, then served as Ai2's COO in December 2025, where she said strong human input helped AI tools align with real-world expectations and teams used feedback for video-language models and vision systems. As of October 2026, she is a Member of Technical Staff at Microsoft and spoke about MAI models at a Microsoft Reactor event in August 2026.
Timeline
-
2012–2013
A Brown PhD and CMU postdoc
She earns a PhD in Cognitive Science at Brown University in 2012; her research examines the neural basis of visual preference for everyday objects. By 2013, she is a postdoctoral fellow at Carnegie Mellon, continuing work with Michael Tarr at the Center for the Neural Basis of Cognition.
-
2014
Co-founding Neon Labs
In 2014, she is Neon Labs' CEO and co-founder, translating visual-preference research into software for choosing online video thumbnails.
-
2015
Neon named a Technology Pioneer
In 2015, the World Economic Forum selects Neon as a Technology Pioneer; Lebrecht is the company's CEO and co-founder.
-
2018–2020
Xnor.ai and Apple's acquisition
In 2018, she is Xnor.ai's director of operations; the company builds software to improve computer vision and aims to track images at 60 frames per second. In early 2020, Apple acquires Xnor.ai after she serves as its senior vice president of strategy and operations.
-
2024
Molmo and PixMo
In 2024, she coauthors Molmo and PixMo, presenting an open family of vision-language models and datasets for detailed image captions, free-form image question answering, and 2D pointing.
-
December 2025
Human feedback at Ai2
In December 2025, she is Ai2's Chief Operating Officer and says strong human input helps AI tools align with real-world expectations. Ai2's Large Language Model and Multimodal Model teams develop video-language models and vision systems that require ongoing human feedback.
-
August 2026
Speaking about Microsoft's MAI models
In August 2026, she speaks at Microsoft's Model Mondays event about the MAI family of models.
-
As of October 2026
Technical staff at Microsoft
She is a Member of Technical Staff at Microsoft.
Career
- Since 2026MicrosoftMember of Technical Staff
- From 2020Apple
- 2018–2020Xnor.aiDirector of operations, Senior vice president of strategy and operations
- 2017–2018Brown University
- From 2014Neon LabsCEO and co-founder
- 2013–2015Carnegie Mellon UniversityPostdoctoral Fellow
- Allen Institute for AI (Ai2)Chief Operating Officer
Education
- 2012Brown UniversityPhD
Sources and credits
Photo: We have no published photo of Sophie Lebrecht yet. Send us a free-licensed photo.
- Research Matters 2014 | Graduate School | Brown University
- Vision - Carnegie Mellon University | CMU
- World Economic Forum Selects CMU BrainHub Startup as Technology Pioneer - News - Carnegie Mellon University
- IoT Cameras Need to be Quick, Clever, and Able to Assign Meaning - IEEE Spectrum
- AI2's Sophie Lebrecht: We need to better understand how generative AI works
- Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models
- How Ai2 builds breakthrough AI faster with quality human data from Prolific
- Model Mondays - Spotlight On: MAI Models | Microsoft Reactor







