Shengzhe Xu, PhD Applied Scientist, Annapurna ML, AWS AI Email: shengzx [at] amazon com Santa Clara, CA Profiles: [Google Scholar] [Orcid]

Research Interests and Background

My research spans synthetic structured data generation, large language model training, agent-based reasoning, and programming languages. I am now committed to building native, large-scale AI model accelerators with my team.

Before joining my current role, I earned my Ph.D. in Computer Science from Virginia Tech, where I was advised by Naren Ramakrishnan. I received my bachelor's degree in Telecommunications Engineering from Beijing University of Posts and Telecommunications. Research and professional experience: Facebook (Ads Core ML: time-series modeling for video clip summarization) [2020], Microsoft Research Asia (Machine Learning Group: multi-agent reinforcement learning) [2018], Google (Input Method Engine) [2015], and Tsinghua University (Tsinghua-Waterloo Joint Research Center for Internet Information Acquisition: topic modeling and semantic analysis of text) [2015].

Publications

• Code Generation and Performance Optimization

  • Can an LLM find its way around a Spreadsheet? Cho-Ting Lee, Andrew Neeser, Shengzhe Xu, Jay Katyan, Patrick Cross, Sharanya Pathakota, Marigold Norman, John C. Simeone, Jaganmohan Chandrasekaran, Naren Ramakrishnan. In Proceedings of the 47th IEEE/ACM International Conference on Software Engineering (ICSE 2025), Ottawa, Canada, April 2025. [pdf]

  • Meditor: inference and application of API migration edits. Shengzhe Xu, Ziqi Dong, Na Meng. In Proceedings of the 27th International Conference on Program Comprehension (ICPC 2019), Montreal, QC, Canada, May 2019. [pdf] [code]

• Synthetic Structured Data Generation

  • Is isotropy a good proxy for generalization in time series forecasting with transformers? Rashed Shelim, Shengzhe Xu, Walid Saad, Naren Ramakrishnan. Transactions on Machine Learning Research (TMLR), Oct 2025. [pdf]

  • The Prompt is Mightier than the Example. Shengzhe Xu, Nikhil Muralidhar, Naren Ramakrishnan. In Proceedings of the ACM SIGKDD Workshop on Prompt Optimization, Toronto, Canada, Aug 2025. [pdf]

  • Why LLMs Are Bad at Synthetic Table Generation (and what to do about it). Shengzhe Xu, Cho-Ting Lee, Mandar Sharma, Raquib Bin Yousuf, Nikhil Muralidhar, Naren Ramakrishnan. In Proceedings of the ACM SIGKDD Workshop on Structured Knowledge for Large Language Models (SKnow-LLM), Toronto, Canada, Aug 2025. [pdf]

  • STAN: Synthetic Network Traffic Generation with Generative Neural Models. Shengzhe Xu, Manish Marwah, Martin Arlitt, Naren Ramakrishnan. In Proceedings of the KDD Workshop on Deployable Machine Learning for Security Defense (MLHat). Virtual Event, Aug 2021. [pdf] [website] [code]

• Large Language Model Optimization

  • Information Guided Regularization for Fine-tuning Language Models. Mandar Sharma, Nikhil Muralidhar, Shengzhe Xu, Raquib Bin Yousuf, Naren Ramakrishnan. In Proceedings of the Conference on Language Modeling (COLM 2024), Philadelphia, PA, Oct 2024. [pdf]

• LLM Agents and Reasoning

  • Utilizing Metadata for better Retrieval-Augmented Generation. Raquib Bin Yousuf, Shengzhe Xu, Mandar Sharma, Andrew Neeser, Chris Latimer, Naren Ramakrishnan. In Proceedings of the 48th European Conference on Information Retrieval (ECIR 2026), Delft, The Netherlands, Mar 2026. [pdf]

  • Can an LLM Induce a Graph? Investigating Memory Drift and Context Length. Raquib Bin Yousuf, Aadyant Khatri, Shengzhe Xu, Mandar Sharma, Naren Ramakrishnan. In Proceedings of the IEEE International Conference on Knowledge Graphs (ICKG 2025), Cyprus, Nov 2025. [pdf]

  • LLM Augmentations to Support Analytical Reasoning over Multiple Documents. Raquib Bin Yousuf, Nicholas Defelice, Mandar Sharma, Shengzhe Xu, Naren Ramakrishnan. In Proceedings of the IEEE International Conference on Big Data (BigData 2024), Washington DC, Dec 2024. Best Paper Award. [pdf]

  • Large Multi-Modal Models (LMMs) as Universal Foundation Models for AI-Native Wireless Systems. Shengzhe Xu, Christo Kurisummoottil Thomas, Omar Hashash, Nikhil Muralidhar, Walid Saad, Naren Ramakrishnan. IEEE Network, July 2024. [pdf]

• Applied Machine Learning

  • Optimizing Product Provenance Verification using Data Valuation Methods. Raquib Bin Yousuf, Hoang Anh Just, Shengzhe Xu, Brian Mayer, Victor Deklerck, Jakub Truszkowski, John C. Simeone, Jade Saunders, Chang-Tien Lu, Ruoxi Jia, Naren Ramakrishnan. In Proceedings of the Innovative Applications of Artificial Intelligence Conference (IAAI 2026), Singapore, Jan 2026. [pdf]

  • Wireless Knowledge Grounding in Smaller LLMs using Retrieval Augmented Generation and Fine-Tuning. Andrew Neeser, Christo Kurisummoottil Thomas, Shengzhe Xu, Naren Ramakrishnan, Walid Saad. In Proceedings of the IEEE International Conference on Communications (ICC 2025), Montreal, Canada, June 2025. [pdf]

  • ML-Assisted Optimization of Securities Lending. Abhinav Prasad, Prakash Arunachalam, Ali Motamedi, Ranjeeta Bhattacharya, Beibei Liu, Hays Mccormick, Shengzhe Xu, Nikhil Muralidhar, Naren Ramakrishnan. In Proceedings of the 4th ACM International Conference on AI in Finance (ICAIF 2023), Brooklyn, NY, Nov 2023. [pdf]

Services