Sripad Karne

I am primarily focused on mechanistic interpretability (SAEs, Probes, CoT, essentially anything model internal related!) and its application for AI Safety. I am interested in cracking open the black box and understanding what's really going on, and that's what my research has focused on.

In particular, I am interested in cross-lingual generalization and how some of the widely used tools and models are affected by this. I believe it's important that AI, and AI Safety, should apply to all, and I think there's still a lot of work that needs to be done here.

On the academic side, I am a Master's student in Data Science @ Columbia University, advised by both Prof. Tianyi Peng and Wenyue Hua. I did my undergrad @ UC San Diego, studying Cognitive Science & ML, advised by Prof. Zhuowen Tu.

Publications

How Far Do Auto-Interpretation Labels Generalize? A Controlled Study Across Languages, Scripts, and Rewordings

Sripad Karne

Submitted to EMNLP 2026 (main track, via ARR)

One Language, Two Scripts: Probing Script-Invariance in LLM Concept Representations

Sripad Karne

Unifying Concept Representation Learning(UCRL) Workshop @ ICLR 2026

The Monitor Knows, the Threshold Doesn’t: Multilingual Miscalibration of Safety Probes

Sripad Karne

Submitted to BlackboxNLP @ EMNLP 2026

Client-Side Optimization for LLM Agent Pipelines: Combinatorial Model Selection with Sample-Efficient Search

Wenyue Hua, Sripad Karne, Qian Xie, Armaan Agrawal, Nikos Pagonas, Wenxin Zhang, Eugene Wu, Kostis Kaffes, and Tianyi Peng

Submitted to NeurIPS 2026 (main track)

A bandit-based approach to assigning heterogeneous models across the steps of a multi-agent pipeline.

AgentOpt Code

Wenyue Hua, Sripad Karne, Qian Xie, Armaan Agrawal, Nikos Pagonas, Kostis Kaffes, and Tianyi Peng

arXiv, 2026

A framework-agnostic optimization layer for client-side model selection in multi-agent LLM systems, released as an installable package.

Machine Learning–Predicted Risk Trajectories for Incident Chronic Kidney Disease and Associations with Post-CKD Outcomes

Aaron Boussina, Amy M. Sitapati, Soo-Young Yoon, Sripad Karne, Jongwoo Seo, Woo-jung Kim, and Hyeon Seok Hwang

Journal of Medical Systems, 2026

Experience

Independent Researcher — Mechanistic Interpretability

Jan 2026 – Present

AI Engineer Intern, IBM

May 2026 – Present · San Francisco

Graduate AI Researcher, DAP Lab, Columbia University

Jan 2026 – May 2026 · Advised by Prof. Tianyi Peng & Dr. Wenyue Hua

Biomedical AI Intern, UC San Diego

May 2025 – August 2025 · Advised by Prof. Aaron Boussina

Sripad Karne

Publications

Experience

Awards

Contact