🧑🚀 🧘 Hello!
Hey there! I'm Gauri, an applied ML and AI scientist currently navigating the exciting frontiers of Generative AI. Think large language models, creative algorithms, and building intelligent systems that learn and adapt.
Featured Publications
CAPTURE: Context-Aware Prompt Injection Testing and Robustness Enhancement
Authors: Gauri Kholkar (First Author), et al.
Conference: Accepted in ACL 2025 LLMSEC Workshop
Prompt injection remains a major security risk for large language models. However, the efficacy of existing guardrail models in context-aware settings remains underexplored, as they often rely on static attack benchmarks. Additionally, they have over-defense tendencies. We introduce CAPTURE, a novel context-aware benchmark assessing both attack detection and over-defense tendencies with minimal in-domain examples. Our experiments reveal that current prompt injection guardrail models suffer from high false negatives in adversarial cases and excessive false positives in benign scenarios, highlighting critical limitations.
What No One Tells You about Securing AI Apps: Demystifying AI Guardrails
Explore how foundational DevSecOps and LLM guardrails offer a clear, actionable path to protecting AI applications from emerging risks.
Volunteering & Community
- ICML 2025 Reviewer
Review long papers for LatinX in AI Workshops (April 2025) - ACL 2025 Reviewer
Review long papers for LLMSec Workshop and GenderBiasNLP Workshop (April 2025) - ICLR 2025 Reviewer
Review long papers for Building Trust in LLMs and LLM Applications ICLR Workshop (Feb 2025) - Pure Storage Empower Me Program mentor
Mentor 3rd year undergraduate student to build LLM application.
Talks & Presentations
I enjoy sharing insights and learnings at conferences and events. You can find details about my upcoming and past talks on the dedicated talks page.
> Opinions expressed are solely my own and do not express the views or opinions of my employer.