Yifu Wu吴一夫

Applied Scientist, AWS Generative AI Innovation Center · Seattle, WA

Email Google Scholar GitHub LinkedIn ORCID

About

I am an Applied Scientist at the AWS Generative AI Innovation Center, where I work on LLM inference efficiency, agent infrastructure, prompt caching and model routing. I care about making models cheaper and faster to serve without giving up correctness, and about measuring that carefully.

Before that, I built multilingual content-safety classifiers and responsible-AI guardrail models at Amazon Alexa, and did clinical NLP research at the University of Colorado Anschutz Medical Campus, contributing to knowledge-graph retrieval and clinical text summarization. I received my Ph.D. from Purdue University, advised by Prof. Jin Wei-Kocsis. My doctoral research covered decentralized and distributed machine learning (including ML systems for NASA spacecraft networks and distributed learning on embedded devices), security for cyber-physical systems, and blockchain-based computing.

Interests: LLM inference and serving · AI agents and evaluation · interpretability · distributed systems and HPC.

Publications

13 peer-reviewed publications. Y. Wu marks me. Full list on Google Scholar.

2026

• LogosKG: Hardware-Optimized Scalable and Interpretable Knowledge Graph Retrieval
  H. Cheng, Y. Wu, S. Khatwani, M. Kruse, D. Dligach, T. Miller, M. Afshar, Y. Gao
  ACL 2026 (Main Conference), 2026. [DOI]

2025

• Large Language Models with Temporal Reasoning for Longitudinal Clinical Summarization and Prediction
  M. Kruse, S. Hu, N. Derby, Y. Wu, et al.
  Findings of ACL: EMNLP 2025, 2025. [DOI]
• Remote Physical Control for Upgrading Heavy Construction Equipment
  M. Soy, J. Fu, Y. Wu, J. Zhang, D. Gan, J. Wei-Kocsis, B.C. Min
  ISARC Proceedings of the International Symposium on Automation and Robotics in Construction, 2025. [DOI]

2022

• A Practical and Stealthy Adversarial Attack for Cyber-Physical Applications
  Y. Wu, J. Wei-Kocsis
  AAAI-22 Workshop on Adversarial Machine Learning and Beyond, 2022.

2020

• DDLPF: A Practical Decentralized Deep Learning Paradigm for Internet-of-Things Applications
  Y. Wu, G.J. Mendis, J. Wei
  IEEE Internet of Things Journal, vol. 8, no. 12, pp. 9740-9752, 2020. [DOI]
• A Blockchain-Powered Decentralized and Secure Computing Paradigm
  G.J. Mendis, Y. Wu, J. Wei, M. Sabounchi, R. Roche
  IEEE Transactions on Emerging Topics in Computing, vol. 9, no. 4, pp. 2201-2222, 2020. [DOI]
• Towards an Adaptive and Attack-Resilient Communication Infrastructures for Smart Grids
  Y. Wu, J. Wei, B.M. Hodge
  Security of Cyber-Physical Systems: Vulnerability and Impact, pp. 293-323, 2020. [DOI]

2019

• A Domain Knowledge-Enabled Hybrid Semi-Supervision Learning Method
  Y. Wu, J. Wei
  IEEE Global Conference on Signal and Information Processing (GlobalSIP), 2019. [DOI]

2017

• Towards Attack-Resilient Communications for Smart Grids with Software-Defined Networking
  Y. Wu, J. Wei
  IEEE Power & Energy Society General Meeting, pp. 1-5, 2017. [DOI]
• A Distributed Middleware Architecture for Attack-Resilient Communications in Smart Grids
  Y. Wu, J. Wei, B.M. Hodge
  IEEE International Conference on Communications (ICC), pp. 1-7, 2017. [DOI]
• A Privacy-Preserving Middleware Mechanism for Smart Grids
  Y. Wu, Y. He, G.J. Mendis, J. Wei
  IEEE 2nd International Conference on Cloud Computing and Big Data Analysis (ICCCBDA), 2017. [DOI]

2016

• An Attack-Resilient Middleware Architecture for Grid Integration of Distributed Energy Resources
  Y. Wu, G.J. Mendis, Y. He, J. Wei, B.M. Hodge
  IEEE International Conference on Internet of Things (iThings) and IEEE Green Computing and Communications, 2016. [DOI]
• Social Community-Based Scheme for Preserving Privacy of Smart Meters
  Q. Gao, Y. Wu, J. Wei
  Proceedings of the Workshop on Communications, Computation and Control for Resilient Smart Energy Systems, 2016. [DOI]

Preprints and thesis

• Real-Time Cascade Mitigation in Power Systems Using Influence Graph Improved by Reinforcement Learning
  K. Zhou, Y. He, C. Zhong, Y. Wu
  arXiv preprint arXiv:2506.08893, 2025. [arXiv]
• Data-driven Computing and Networking Solution for Securing Cyber-Physical Systems
  Y. Wu
  Ph.D. Dissertation, Purdue University, 2024. [thesis]

Patents

• Blockchain-Empowered Crowdsourced Computing System
  J. Kocsis, Y. Wu, G.J.M.I. Liyangahawatte
  US Patent 11,063,759, granted 2021.
• 3S-Chain: Smart, Secure, and Software-Defined Networking (SDN)-Powered Blockchain-Powered Networking and Monitoring System
  J. Kocsis, M.P.S. Fernando, Y. Wu
  US Patent 12,034,770, granted 2024.

Experience

• Applied Scientist, AWS Generative AI Innovation CenterApr 2026 – present
• Applied Scientist, Amazon AlexaApr 2025 – Apr 2026
• NLP Data Scientist, University of Colorado Anschutz Medical CampusAug 2024 – Mar 2025
• Graduate Research Assistant, Purdue UniversityDec 2019 – May 2024
• Graduate Research Assistant, University of Akron and Iowa State UniversityAug 2015 – Jul 2019

Education

• Ph.D., Computer and Information Technology, Purdue University2024
• M.E., VLSI Systems, University of Limerick2015
• B.Eng., Automation, Harbin Institute of Technology2011

Service

Reviewer for ICLR, ACL Rolling Review (ARR), IEEE Internet of Things Journal, and IEEE Transactions on Consumer Electronics.