Profile banner
Vrushank Gunjur

Vrushank Gunjur

Stanford CS student and researcher — systems, security, and ML

vrushank
Stanford, California
Joined October 2025
Favicon imageprofiles.stanford.edu
Favicon imagescholar.google.com
Favicon imagecommencement.stanford.edu

Network

876 connections
🎓
Stanford Engineering & CS Alumni
🔬
Stanford STEM Academics & Researchers
🚀
YC AI Startup Founders
🤖
AI & Tech Innovators
💻
Stanford & MIT AI/CS Professionals
📈
Citadel Securities Quant Finance
🧬
Stnfrd BioEng

Summary

Applied systems and security researcher — hands-on internet measurement and security work. Vrushank has been listed as a member/contributor in Stanford’s Empirical Security Research Group and is involved in projects that build measurement systems and analyze real-world security behavior. stanford+1
ML / model-inference and systems-oriented research. Coauthored work on speculative decoding (entropy‑informed dynamic tree shaping) that targets faster LLM inference, demonstrating experience at the intersection of model-research and systems engineering. ajwaitz
Privacy, cryptography, and blockchain governance research. Coauthor on Kite, a protocol for private delegation of voting power for DAOs, showing interest and contributions in privacy-preserving protocol design and blockchain governance. arxiv
Practical/engineering and teaching experience — participates in hackathons and builds engineering projects (e.g., Plato), and has supported teaching and research roles at Stanford (research assistant / section leader roles). This mix indicates strengths in shipping prototypes, mentoring/teaching, and contributing to multidisciplinary research teams. devpost+2

Work

Education

Projects

Writing

Kite: How to Delegate Voting Power Privately

January 1, 2025

ArXiv paper introducing Kite, a protocol for private delegation of voting power (applicable to DAOs), including a UC-style security analysis and an Ethereum (Governor Bravo) implementation evaluation.

Favicon imagearxiv.org

Fast Inference with Dynamic Tree Speculative Decoding

January 1, 2023

Research paper describing an entropy-informed technique to shape speculative decoding trees for faster LLM inference; demonstrates practical generation speedups through dynamic tree shaping.

Favicon imageajwaitz.org