Amrit Puhan

I am a Software Development Engineer II in the Applied AI Solutions org at Amazon Web Services, where I build AI systems over petabyte-scale automotive and industrial data. For autonomous driving (ADAS/AV) teams, I work on a multi-modal video search system that helps developers surface rare edge cases across video, sensor and annotation data using plain-English queries. In the industrial space, I work on time series anomaly detection for predictive maintenance of equipment.

I completed my MS in Informatics (Data Science concentration) at Penn State in the FAIR Lab, advised by Dr. Hadi Hosseini. I was also mentored by and collaborated with Dr. Debmalya Mandal. My thesis, Recovering Ground Truth Rankings When the Majority Is Misinformed, won Best Master’s Thesis on an AI-related topic at the Penn State AI Awards. This work on surprisingly popular voting led to publications at NeurIPS 2024, WWW 2025 and KDD 2026. Before that, I earned my B.Tech in Computer Science and Engineering from NIT Rourkela, where I did my undergraduate thesis in the Intelligent Computing and Computer Vision group with Dr. Anup Nandy.

my career at a glance

More broadly, I am interested in:

• Computational social choice: recovering ground truth from noisy, disagreeing preferences through rank aggregation and surprisingly popular voting
• Aligning AI with human preferences: preference elicitation and learning from human feedback for large language models
• Multi-agent LLM systems, including inference-time preference elicitation and multi-agent fine-tuning
• Probabilistic and Bayesian models of human behavior, such as Mallows and Plackett-Luce ranking models
• Multi-modal search and retrieval over large-scale video and sensor data

academic service

• Reviewer, KDD 2026, Datasets and Benchmarks Track
• Reviewer, The ACM Web Conference (WWW) 2025, Main Track