Aashiq Muhamed

I am a Ph.D. student in the CMU Machine Learning Department, advised by Professors Mona Diab and Virginia Smith. I also closely collaborate with Professor Andrew Ilyas.

My research develops the science and safety infrastructure for autonomous agentic systems. I work on both the foundational interpretability methods needed to understand how models reason internally, and the applied tools to monitor and govern agents in deployment. I’m particularly interested in:

• Foundations of Mechanistic Interpretability: Developing SAEs, circuit analysis, and interpretability methods to understand model internals—including feature consistency, rare concept detection, unlearning, and the geometry of learned representations.
• Interpretability of Reasoning Models: Understanding how reasoning capabilities emerge and can be intervened on during training across paradigms (SFT, RLVR, and beyond), and what drives robustness in finetuned reasoning models.
• Agent Safety Monitoring and Evaluation: What makes an effective safety monitor? I study chain-of-thought monitoring, mechanistic monitors, and benchmarking their robustness against direct misuse and evasion.
• Multi-Agent Collusion and Coordination: Using mechanistic tools to detect and understand collusion, deception, and covert coordination in multi-agent systems.
• Tamper Resistance and Open-Weight Safety: Making safety properties robust in open-weight models where adversaries have full access to modify weights.

I am supported by the Amazon PhD Fellowship, Anthropic Fellows Program, the Siebel Scholars Foundation, the Cooperative AI PhD Fellowship, Longview Philanthropy, CAIS, SPAR, and MATS.

background

I bring an interdisciplinary perspective, with degrees spanning engineering, language technologies, and machine learning. I earned my B.Tech in Mechanical Engineering from the Indian Institute of Technology, Roorkee, where I was awarded the President’s Gold Medal. I subsequently completed an MS in Mechanical Engineering at Stanford University and an MS in Language Technologies from the CMU Language Technologies Institute.

Before beginning my Ph.D., I spent five years in industry as an Applied Scientist at Amazon, working across diverse AI applications including AWS DeepComposer (2019-2021), Amazon Search M5 (2021-2022), and AWS AI (2022-2023). This industry experience deeply informs my research, emphasizing systems that are both theoretically grounded and practically deployable.

I’m always open to collaboration — feel free to reach out.