whoami

Bhuvan Chandra Koduru

Pittsburgh, PA CMU LTI Simular.ai

I'm a Master's student in AI/NLP at Carnegie Mellon University's Language Technologies Institute, working at the intersection of AI safety, multimodal learning, and agentic systems.

Currently at Simular.ai reverse-engineering vision-language models for real-time intervention on computer-use agents. Previously built evaluation harnesses simulating 8,000+ multi-agent games at CMU's Robotics Institute.

Before CMU, I architected multi-agent AI systems at Adobe, reducing documentation time by 70% and advancing to top-24 semifinalists worldwide in an innovation challenge.

Bhuvan Koduru

01 Recent News

Jun 2026
Joined Simular.ai as Research Intern—working on AI safety and computer-use agents with Xin Eric Wang, Jiachen Yang, and Ang Li.
May 2026
4 papers accepted! CoLM 2026, ICML MI Workshop, INTERSPEECH 2026, and CoLM Lifelong Agents Workshop.
Jan 2026
Started as Graduate Research Assistant at CMU Robotics Institute—building evaluation harnesses for multi-agent LLM deception.
Aug 2025
Joined Carnegie Mellon University's Language Technologies Institute for MS in AI/NLP.

02 Selected Publications

Heard but Not Heeded: Paralinguistic Information Encoding and Loss in Audio-Language Models
B. Koduru, D. Alharthi, R. Singh, B. Raj
CoLM 2026
AmongUs-X: Benchmarking Strategic Deception of LLM Agents via Theory-of-Mind Metrics
K. Mirakhor*, B. Koduru*, A. Kontogiannis, R. Zabounidis, K. Sycara, W. Kim
NeurIPS 2026 under review
Subliminal Prosody Learning: Auxiliary Emotion Supervision Redistributes Affective Representations
B. Koduru, D. Alharthi, R. Singh, B. Raj
ICML MI 2026
$ ls ~/publications