Tushar Karayil
Researcher/Scientist
I'm a researcher working at the intersection of Machine Learning, Large Language Models, and AI Safety. My work explores how machines learn, reason, and align — grounded in optimization theory and informed by philosophy. This is a space where I share ideas, research, and reflections on intelligence from both mathematical and philosophical perspectives.
CV Highlights
Full CV →Career Summary
- 17+ years of industry experience.
- 12+ years of experience in Deep Learning research.
- 10+ years of experience in Software Development.
Education
PhD (Dr. rer. nat.), Deep Learning based Affective Image Captioning
DFKI (German Research Center for AI)/RPTU · Germany
2016 – 2020
Current Position
Research Scientist
ZScaler · USA / India
2023 – Present
Previous Position
Senior Researcher
DFKI (German Research Center for AI) · Germany
2016 – 2023
Recent Publications
View all →Moral Preferences of LLMs Under Directed Contextual Influence
P. Blandfort, T. Karayil, U. Pawar, R. Graham, A. McKenzie, D. Krasheninnikov · arXiv preprint
Moral benchmarks for LLMs typically use context-free prompts, implicitly assuming stable preferences. We study how directed contextual influences resh...
Rethinking Software Design with Large Language Models Intelligent Interfaces
C. N. Coelho, H. Xiong, T. Karayil, S. Koratala, R. Shang, J. Bollinger, M. Shabar, S. Nair · International Conference on Multimodal Interaction (ICMI)
The advancement of Large Language Models (LLMs) has led to a rapid expansion of their applications, including in software design. We propose a new app...