Arth Singh
Research Engineer, AI Safety at AIM Intelligence.
Upcoming OMSCS at Georgia Tech Fall 2026.
News
- Apr 2026Submitted SENTINEL to ICML AI4Good Workshop 2026
- Apr 2026Submitted Readability vs. Controllability Gap to ICML Mechanistic Interpretability Workshop 2026
- Mar 2026VLM compression paper published on arXiv
- Mar 2026Submitted Re-Mask and Redirect to ACL TrustNLP 2026
- Feb 2026Started SPAR Spring 2026 fellowship — scalable oversight with Google DeepMind
The Journey
Georgia Tech OMSCS
Incoming Master's student in Computer Science (Machine Learning specialization) at Georgia Institute of Technology.
Georgia Tech OMSCS
Incoming Master's student in Computer Science (Machine Learning specialization) at Georgia Institute of Technology.
SPAR Spring 2026 — Scalable Oversight
Second SPAR fellowship working on scalable oversight research with Rishub Jain (Google DeepMind). Investigating how to maintain human control over increasingly capable AI systems.
SPAR Spring 2026 — Scalable Oversight
Second SPAR fellowship working on scalable oversight research with Rishub Jain (Google DeepMind). Investigating how to maintain human control over increasingly capable AI systems.
SNU PI Lab — Adversarial Attacks on GUI Agents
Research at Seoul National University's PI Lab with Prof. Youngjae Yu. Developed adversarial attacks targeting Mobile GUI agents — the GART framework.
SNU PI Lab — Adversarial Attacks on GUI Agents
Research at Seoul National University's PI Lab with Prof. Youngjae Yu. Developed adversarial attacks targeting Mobile GUI agents — the GART framework.
AIM Intelligence, Seoul
Joined as a Research Engineer in the AI Safety department. Lead adversarial red-teaming engagements for enterprise clients including BMW, Samsung, Deutsche Telekom, and Glean. Additionally working on Physical AI Safety, discovering vulnerabilities in robotic systems alongside LG.
AIM Intelligence, Seoul
Joined as a Research Engineer in the AI Safety department. Lead adversarial red-teaming engagements for enterprise clients including BMW, Samsung, Deutsche Telekom, and Glean. Additionally working on Physical AI Safety, discovering vulnerabilities in robotic systems alongside LG.
SPAR Fellow at FAR.AI
Built an automated red-teaming pipeline for frontier models as a SPAR Fall 2025 fellow. Developed scalable attack methodologies for evaluating model safety.
SPAR Fellow at FAR.AI
Built an automated red-teaming pipeline for frontier models as a SPAR Fall 2025 fellow. Developed scalable attack methodologies for evaluating model safety.
Anthropic Constitutional Classifiers (HackerOne)
Invited by Anthropic to test Constitutional Classifiers — their safety layer for Claude. Identified bypass techniques through HackerOne's private program.
Anthropic Constitutional Classifiers (HackerOne)
Invited by Anthropic to test Constitutional Classifiers — their safety layer for Claude. Identified bypass techniques through HackerOne's private program.
Mozilla Project Ødin — Peaked World #2
Red-teamed frontier AI models through Mozilla's bug bounty program. Peaked at #2 on the global leaderboard, earning $15k+ in bounties for discovering critical safety vulnerabilities.
Mozilla Project Ødin — Peaked World #2
Red-teamed frontier AI models through Mozilla's bug bounty program. Peaked at #2 on the global leaderboard, earning $15k+ in bounties for discovering critical safety vulnerabilities.
AI Safety Camp — EU AI Policy
Participated in AI Safety Camp working on EU AI policy research, exploring regulatory frameworks for safe AI deployment.
AI Safety Camp — EU AI Policy
Participated in AI Safety Camp working on EU AI policy research, exploring regulatory frameworks for safe AI deployment.
NIT Agartala — B.Tech Biotechnology
Started undergraduate studies in Biotechnology at National Institute of Technology, Agartala.
NIT Agartala — B.Tech Biotechnology
Started undergraduate studies in Biotechnology at National Institute of Technology, Agartala.