Research
Interpretable vision systems, representation analysis, and responsible AI.
I work on interpretable AI systems, computer vision, learned feature representations, category structure, generalization, and responsible LLM use.
Research Profile
I study how computer vision models encode category structure, how internal representations can be inspected and visualized, and how responsible AI practices can reduce risks in model use and learning.
Computer VisionLearned FeaturesGeneralizationXAILLM Safety
Boston University Computational Neuroscience & Vision Lab
AI/ML Researcher
2024 - Present
- Designed and implemented CNN pipelines using MATLAB's Statistics and Machine Learning Toolbox.
- Investigated how learned feature representations encode category structure and generalization in computer-vision tasks.
- Analyzed internal network representations using dimensionality reduction.
- Designed search algorithms to identify patterns within learned feature spaces.
- Integrated visualization techniques to validate relationships between attention patterns and feature-space organization.
- Optimized model performance through systematic hyperparameter tuning.
Publications
Published and accepted research.
Object Detection, Recognition, Deep Learning, and the Universal Law of Generalization
Faris B. Rustom, Rohan Sharma, Haluk Öğmen, Arash Yazdanbakhsh
Neural Computation, MIT Press
2026
DOI 10.1162/NECO.a.1483Large language model dependence and academic burnout: Cognitive and behavioral effects in undergraduate learning
Rohan Sharma, Diane Mello-Goldner
Poster accepted for presentation at the Fall 2026 Annual Meeting of the New England Psychological Association, Salem State University