Mariya I. Vasileva

Senior Research Scientist, Meta Superintelligence Labs, Facebook

  • Joined July 2019

Names

Mariya I. Vasileva

Emails

****@illinois.edu (Confirmed)
,
****@gmail.com (Confirmed)
,
****@amazon.com (Confirmed)
,
****@meta.com (Confirmed)

Career & Education History

Senior Research Scientist
Meta Superintelligence Labs, Facebook (meta.com)
2025Present
 
Head of Research
Kaiber (kaiber.ai)
20232023
 
Applied Scientist
AWS (amazon.com)
20202023
 
PhD student
University of Illinois, Urbana Champaign (illinois.edu)
20162020
 
Undergrad student
California Institute of Technology (caltech.edu)
20092013
 

Advisors, Relations & Conflicts

Coauthor
Present
 
Coauthor
Present
 
Coauthor
Present
 
Coauthor
Present
 
Coauthor
Present
 
Social
Present
 
PhD Advisor
20162020
 

Expertise

Mechanistic interpretability
2025Present
 
LLM-as-a-judge systems
2025Present
 
Scalable oversight
2024Present
 
Agentic systems
2024Present
 
Foundation and frontier model scaled evaluations
2024Present
 
Video understanding and analysis
2023Present
 
Video generation
2023Present
 
Image-to-video generation
2023Present
 
Text-to-video generation
2023Present
 
Text-to-image generation
2023Present
 
Temporal consistency and motion representation
2023Present
 
Diffusion models
2022Present
 
AI Safety, Trust & Alignment
2022Present
 
Multimodal safety and robustness
2022Present
 
Robustness, accountability and transparency
2021Present
 
Active learning, human-in-the-loop
2021Present
 
Fairness
2020Present
 
Foundation model posttraining
2020Present
 
Contrastive learning
2018Present
 
Interpretability of black-box models
2018Present
 
Explainability
2017Present
 
Visual search and retrieval
2017Present
 
AI for fashion applications
2017Present
 
Generative models
2017Present
 
Vision and language models
2017Present
 
Transformer models
2017Present
 
Semi-supervised and unsupervised learning
2016Present
 
Multimodal understanding and generation
2016Present
 
Self-supervised learning
2016Present
 
Few-shot and zero-shot learning
2015Present
 
Few- and zero-shot learning
2015Present
 
Machine learning
2014Present
 
Representation learning
2014Present
 
Deep learning
2014Present
 
Embedding methods
2014Present
 
Metric learning
2014Present
 
Computer vision
2012Present