Mariya Vasileva
Sr. Research Scientist @ Meta Superintelligence Labs. My experience lies at the crossover between generative models, multimodal learning, vision and language, foundation models, safety and alignment.
- Role
- Senior Research Scientist at Meta
- Location
- New York, NY, US
- LinkedIn followers
- 500 followers
About Mariya Vasileva
My name is Mariya, and I work at the intersection of generative models, multimodal learning, vision–language systems, and safety and alignment. My research focuses on vision and language learning in large-scale multimodal models, with an emphasis on visual reasoning and evaluation science for frontier systems. In particular, my work tightly integrates vision–language model development — representation learning, modality alignment, posttraining, and synthetic data — with benchmark design, interpretability, and capability, alignment, and safety evaluations. A unifying theme of my work is multimodal safety, trust, and alignment, studied empirically through scalable oversight, stress-testing, and robustness analysis across the full model lifecycle.I spent a number of years in industry research and worked on a range of problems, from image and video generative models, to visual recommender engines, to 2D-to-3D human body shape and pose modelling, to synthetic data generation for efficient training of foundation models at scale, to fairness and explainability of AI systems. I obtained my PhD in Computer Science from the University of Illinois at Urbana-Champaign under the advisorship of professor David A. Forsyth, where I researched problems in vision and language, visual search and retrieval, and applications of computer vision in the fashion domain.
Experience
Senior Research Scientist
Mar 2025 — Present · New York, NY, US
Designed and implemented rigorous, large-scale evaluation frameworks for AI trust, safety, and alignment with a focus on systematic benchmarking, quantitative risk assessment, and scalable oversight of multimodal foundation and frontier models across high-impact domains- Developed core components of end-to-end scalable evaluation ecosystems spanning policy operationalization, large-scale vision data sourcing and annotation workflows, LLM/VLLM-as-a-judge framework development and calibration against human baselines, and risk metric design and instrumentation, enabling continuous oversight of model behavior in close collaboration with cross-functional partners; these evaluation frameworks now underpin all MSL pre-release model and product testing
Education
Caltech
Bachelor of Science (BS), Mechanical Engineering
2009 — 2013
University of Illinois Urbana-Champaign
Doctor of Philosophy (Ph.D.), Computer Science
Caltech
Bachelor of Science (BS), Business Economics and Management
2009 — 2013
Caltech
Minor, Control and Dynamical Systems
2009 — 2013
Skills
- Signal Processing
- Mathematica
- Finite Element Analysis
- R
- Caffe
- Java
- Pro Engineer
- Matlab
- Weka
- Nltk
- Hadoop
- Deep Learning
- Mathematical Modeling
- Knowledge Representation
- Data Analysis
- Product Design
- Sql
- Labview
- Simulations
- Statistics
- Torch
- Solidworks
- Machine Learning
- Combinatorics
- Optimization
- Matconvnet
- Numerical Analysis
- Algorithms
- Tableau
- Graph Theory
- Mobile Robotics
- Ansys
- Control Systems Design
- Artificial Intelligence
- Robotics
- Apache Spark
- Computational Photography
- Computer Vision
- Geometric Dimensioning & Tolerancing
- Guidance Navigation & Control
Find verified contacts for anyone on LinkedIn
Unifers gives sales teams verified emails and direct dials, enriched profiles, and outreach that lands in the inbox.
Free plan included · No credit card required
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.