Kartikeya Chitranshi

Machine Learning Researcher specializing in adversarial robustness and multimodal AI systems, with an MSc in Scientific Computing from TU Berlin and 2+ years of research experience at Zuse Institute Berlin on foundation model security.

My work spans publishing on counterfactual explanations, developing production ML solutions, and deploying robust systems using PyTorch, LangChain, LangGraph and HuggingFace. I thrive at the intersection of research and engineering, seeking ML/AI Engineer roles to build secure, scalable AI with real-world impact.

I have over 3 years of ML research experience, currently working as a Student Research Assistant at Zuse Institute Berlin (2023-2025) where I specialize in Vision-Language Model security, implementing supervised fine-tuning of CLIP models and developing adversarial attack methodologies using PyTorch on multi-GPU infrastructure. My career began with foundational industry experience at TestAIng.com (2021) implementing adversarial attack algorithms on deep neural networks, and at Peacock Solar (2020) where I developed predictive ML models achieving 85%+ accuracy for market forecasting across 28 Indian states. My expertise has evolved from traditional ML applications to cutting-edge AI research in multimodal systems, adversarial training, and RAG architectures, culminating in academic publications on spurious correlation removal and a Master's thesis on multi-modal foundation model robustness, positioning you as a specialist in AI safety with both theoretical depth and practical implementation skills across distributed computing environments.

My most relevant education consists of my completed Master's degree in Scientific Computing at TU Berlin (2022-2025), which directly aligns with my ML research career through specialized coursework in Numerical Linear Algebra, Machine Learning, Deep Learning, Machine Intelligence, Modern Algorithms for Multiagent Learning, Optimization under Uncertainty, High-Dimensional Optimization and Learning, and Reinforcement Learning. This advanced program provides the theoretical foundation for my current work in Vision-Language Models and adversarial AI systems. My Bachelor's in Physical Sciences from University of Delhi (2018-2021) established your quantitative foundation with courses in Numerical Methods, Calculus and Matrices, Differential Equations, and programming fundamentals including Java and system architecture, which prepared me for the transition into advanced computational research. The Master's program is particularly relevant as it directly supports my current research focus on multimodal AI robustness and my thesis work on Vision-Language Model security at Zuse Institute Berlin.