Alireza Bagheri Garakani

I’m a Senior Applied Scientist at Amazon, where I’ve spent the last decade building and launching production ML systems for search and recommendations serving billions of queries. My current research focuses on reinforcement learning for large language models — policy optimization, token-level credit assignment, and reward design — alongside leading price personalization work adopted worldwide across product search. My work has appeared at venues including NeurIPS, KDD, WWW, JMLR, and ICASSP.

Before Amazon, I was a research assistant at the University of Southern California, advised by Fei Sha, working on large-scale kernel methods and distributed optimization, and spent a year at the University of Washington hosted by Carlos Guestrin. Earlier, I was a software engineer at Cisco Systems working on language modeling for speech recognition. I received my BS from UW and MS from USC, both in Computer Science. [linkedin] [google scholar] [github]

Select Publications:

• [Under Review, NeurIPS 2026] Attribution-Guided Query Generation for Retrieval (AGQG)
• [Under Review, NeurIPS 2026] Owen-Shapley Policy Optimization (OSPO): A Principled RL Algorithm for Generative Search LLMs [Paper]
• [NeurIPS 2025 Workshop] Large Language Model-Enhanced Reinforcement Learning for Diverse and Novel Recommendations [Paper]
• [KDD 2025] Learning Attribute as Explicit Relation for Sequential Recommendation [Paper]
• [ECNLP 2022] Improving Relevance Quality in Product Search using High-Precision Query-Product Semantic Similarity [Paper]
• [ECNLP 2022] Spelling Correction using Phonetics in E-commerce Search [Paper]
• [WWW 2020] Multi-objective Ranking via Constrained Optimization [Paper]
• [JMLR 2019] Kernel Approximation Methods for Speech Recognition [Paper]
• [ICASSP 2016] A Comparison Between Deep Neural Nets and Kernel Acoustic Models for Speech Recognition [Paper]
• [SDM 2015] A Distributed Frank-Wolfe Algorithm for Communication-Efficient Sparse Learning [Paper] [Slides] [Code]