Scholar
Saeed Vahidian
Google Scholar ID: 8Jd1aUEAAAAJ
University of California, San Diego (UCSD), Duke University
Citations & Impact
Citations
856
H-index
12
i10-index
15
Publications
20
Co-authors
13
list available
Contact
Publications
Resume (English only)
Academic Achievements
- Jan 2025: One paper accepted at ICLR.
- Oct 2024: Two papers submitted to ICLR.
- Sep 2024: One paper submitted to Journal of Machine Learning Research (JMLR).
- Jun 2024: Two papers accepted at CVPR.
- Jun 2024: Chair at CVPR 2024; organized the 1st Workshop on Dataset Distillation for Computer Vision.
- Jun 2024: Co-organized the 3rd FedVision Workshop at CVPR.
- Feb 2024: One paper accepted at ECCV.
- Research supported by major grants including:
- - $20.39M AI Institute for Edge Computing Leveraging Next Generation Networks (Athena);
- - $600K National Science Foundation (NSF) Grant.
- Notable publications include:
- - 'CoreInfer: Accelerating Large Language Model Inference with Semantics-Inspired Adaptive Sparse Activation' (2024, arXiv);
- - 'Group Distributionally Robust Dataset Distillation with Risk Minimization' (ICLR 2025).
Research Experience
- Postdoctoral Scholar, Duke University, April 2023 – Present.
- PhD Student, UC San Diego, Sep 2018 – Apr 2023.
- ML Researcher (Summer Intern), Qualcomm, Summer 2021.
- Invited Collaboration on DDF project, NASA, 2022.
- Research Collaboration, Stanford University, 2018–2019.
- Research Collaboration, McGill University, 2014–2016.
Background
- Currently a postdoctoral researcher at Duke University, working with Prof. Yiran Chen.
- Research focuses on Multimodal Synthetic Data Generation (Vision ↔ Language ↔ Audio) to advance how foundation models learn across modalities.
- Designs high-fidelity, controllable synthetic datasets to enable robust training pipelines for next-generation multimodal AI systems.
- Work is built upon three core pillars: Controllability (fine-grained control over generated data), Explainability (understanding why and how synthetic data is created), and Robust Learning on Synthetic Data (enhancing model generalization and robustness).
- Also explores Edge Intelligence for privacy-aware synthetic data generation and federated training under computational and privacy constraints.
