About Me
Research taste, engineering discipline, and an interdisciplinary lens.
I work at the intersection of AI/ML research engineering, computer vision, computational neuroscience, and trustworthy AI. I am especially interested in how learned systems represent information, generalize across categories, and can be made more interpretable, private, and safe.
Profile
Rohan Sharma
I am an AI/ML research engineer focused on interpretable and trustworthy AI systems, with work spanning computer vision, deep-learning representation analysis, feature-space visualization, and model interpretability. My background connects computer science, neuroscience, and psychology, and I am pursuing an MSE in Artificial Intelligence at the University of Pennsylvania.
My work combines computer science, neuroscience, and psychology to study how intelligent systems learn, organize information, and generalize. That blend is the lens I bring to AI labs and research engineering teams building systems that need to be both capable and understandable.
Alongside research, my Collins Aerospace experience gives me practical software and systems engineering grounding: verification, validation, integration, automated tests, requirements, and cross-team coordination in safety-critical environments.
AI/ML Research Engineering
I work on research problems that require both modeling intuition and reliable implementation.
Computer Vision
My lab work investigates learned feature representations, category structure, and generalization in vision tasks.
Interpretability
I focus on representation analysis, feature-space visualization, dimensionality reduction, and mechanistic interpretability.
Trustworthy AI
My interests include privacy, safety, responsible AI, and the practical risks around LLM dependence.
LLM Safety And Tuning
I bring research interest in responsible language model use, tuning behavior, safety, and privacy.
Software Systems
My systems engineering experience spans requirements, integration, automated testing, and safety-critical avionics contexts.