Saranya Venkatraman
Applied Scientist · Amazon
I'm an Applied Scientist at Amazon, where I build and evaluate agentic AI systems. My current work focuses on user simulation: building agentic "twins" of users. I completed my PhD at Penn State University, advised by Prof. Dongwon Lee in the PIKE research group, and was a Visiting Scholar at New York University in Prof. He He's group. My PhD research focused on machine-generated text detection, authorship attribution, and obfuscation for Large Language Models (LLMs), and on what does or doesn't make LLM-generated text "human-like."
Along the way, I was a research intern at Google, Samsung, and Cadence, building RL, NLP, and ML solutions for dialogue-based recommenders, voice assistants, and hardware design layouts. I'm broadly curious about cognitively plausible architectures for language applications, and always happy to connect.
Selected publications
- AgenTwin: An End-to-End Framework for Building Agentic Twins of Mobile Users
Saranya Venkatraman, Thanh Tran, Shunyan Luo, Devin Chen, Yuning Wu, Kai Wei, Allie Colin, Guido Imbens, Ido Rosen, Christine Ambrose
- CollabStory: Multi-LLM Collaborative Story Generation and Authorship Analysis
Saranya Venkatraman, Nafis Irtiza Tripto, Dongwon Lee
- GPT-who: An Information Density-based Machine-Generated Text Detector
Saranya Venkatraman, Adaku Uchendu, Dongwon Lee
- A Ship of Theseus: Curious Cases of Paraphrasing in LLM-Generated Texts
Nafis Irtiza Tripto, Saranya Venkatraman, Dominik Macko, Robert Moro, Ivan Srba, Adaku Uchendu, Thai Le, Dongwon Lee
- ALISON: Fast and Effective Stylometric Authorship Obfuscation
Eric Xing, Saranya Venkatraman, Thai Le, Dongwon Lee
- ClassInSight: Designing Conversation Support Tools to Visualize Classroom Discussion for Personalized Teacher Professional Development
Tricia J Ngoon, S Sushil, Angela Stewart, Ung-Sang Lee, Saranya Venkatraman, Neil Thawani, Prasenjit Mitra, Sherice Clarke, John Zimmerman, Amy Ogan
Experience
- Applied Scientist, Amazon
Building and evaluating agentic AI systems, with a focus on user simulation.
- Research Intern, Google (Google Assistant)
Built a recommender dialogue agent using hierarchical Soft Actor-Critic for a hybrid action space.
- Research Intern, Samsung Research America
Built an NLU intent-to-action service for Bixby using semantic similarity and word embeddings.
- Machine Learning Intern, Cadence Design Systems
Built ML pipelines for proprietary chip layout images and a hierarchical clustering prototype.
Tutorials
Misc.
Outside work, I run, snowboard, and collect vinyl records. In another life, I sang Indian Classical music and a cappella.
