Scholar
Fuxun Yu
Google Scholar ID: t8vayXEAAAAJ
Principal Research Manager, Microsoft
Citations & Impact
Citations
1,685
H-index
20
i10-index
30
Publications
20
Co-authors
0
Publications
Resume (English only)
Academic Achievements
- - Team won Top-1 in EarthVision Embed2Scale Geo-Embedding Challenge at CVPR'25
- - Collaborative work on High-order Neural Network Design accepted in NeurIPS'24
- - Collaborative work on Federated Learning accepted in ACSAC'24
- - Collaborative work on LLM-inspired Retrieval-Augmented Detector Adaptation accepted in ECCV'24 HCV Workshop
- - Collaborative work on Out-of-Distribution Detection accepted in ICML'24
- - Collaborative work on High-Order Neural Network Design received Best Paper Award Nomination in ASP-DAC'24
- - Collaborative work on Multi-Tenant Federated Learning received Best Paper Award in MLSys CrossFL Workshop'22
- - Collaborative work on Quadratic Neural Networks received Outstanding Paper Award in MLSys'22 (Top 5 paper)
- - Received Outstanding Academic Achievement Award 2022 of GMU Volgenau School of Engineering
- - Work on memory optimization for recommendation model (during Facebook Internship) accepted in ICDCS'22
- - Poster on GPU-aware DNN design accepted in EuroSys'22
- - Work on multi-tenant DNN scheduling on GPUs accepted in ICCAD'21
- - Work on feature-aligned federated learning (Fed^2) accepted in KDD'21
- - Project (Privacy-preserving FL with personal mobility data) won the first prize in IEEE Services Hackathon 2020
- - Work 'Antidote' on dynamic feature pruning received Best Paper Award Nomination in DATE'20
- - Work 'Interpreting and Evaluating Adversarial Robustness' accepted in IJCAI'19
- - Work on Gradient-Free DNN Training using ADMM by Junxiang and Fuxun accepted in KDD'19
Research Experience
- - Principal Research Manager at Microsoft
- - Interned at Facebook (Infrastructure Team, Capacity Engineering & Analysis) in summer 2021, working on memory optimization for recommendation models
- - Interned at Microsoft Research (Redmond) under the supervision of Dr. Di Wang in summers 2019 and 2020
- - Participated in multiple collaborative research projects covering a wide range of topics such as high-order neural network design, federated learning, LLM-inspired retrieval-augmented detector adaptation, etc.
Background
- Research interests include multi-modal geospatial foundational model, cross-modality vision language modeling, VLM finetuning and reinforcement learning, auto-regressive modeling for graphs, generative artificial intelligence (Gen AI), agentic AI workflow development, retrieval-augmented learning and adaptation, stable diffusion for downstream CV task, high-performance deep learning systems, full-stack optimization on GPUs (algorithm/compiler/runtime), recommendation model memory system optimization, interpretable and explainable artificial intelligence, DNN security and adversarial robustness.
Co-authors