Chongyang Shi
Scholar

Chongyang Shi

Google Scholar ID: xNLhF2UAAAAJ
Google DeepMind
ML Security and Privacy
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
2,869
 
H-index
5
 
i10-index
4
 
Publications
7
 
Co-authors
0
 
Contact
Emailhello@chongya.ngGitHubOpen ↗LinkedInOpen ↗
Publications
7 items
C-IDS: Solving Contextual POMDP via Information-Directed Objective
2026
Cited
0
Lessons from Defending Gemini Against Indirect Prompt Injections
2025
Cited
0
Defeating Prompt Injections by Design
2025
Cited
0
Large Language Models Can Verbatim Reproduce Long Malicious Sequences
2025
Cited
0
Deep Robust Reversible Watermarking
2025
Cited
0
Synthesis of Dynamic Masks for Information-Theoretic Opacity in Stochastic Systems
2025
Cited
0
Retrieval-Augmented Generation by Evidence Retroactivity in LLMs
2025
Cited
0
Resume (English only)
Background
  • Currently a Staff Engineer at Google DeepMind, leading Gemini's adversarial security evaluations and post-training.
Miscellany
  • Works on a range of personal and open-source projects related to security, DevOps, or just hobby; can be found via GitHub. Also operates a personal, cost-optimised Kubernetes cluster to host many services.
Co-authors
0 total
Co-authors: 0 (list not available)