NEURAL FLOW · ABOUT ME
~/wonderland — zsh ▢ ▢ ◯

About Me

$ Bio of a curious mind — research, travels, and the occasional detour.
一个好奇的人——研究、旅行、以及偶尔的岔路。

About me

My name is Xupeng Chen, born in CHINA in 1995.

I’m currently a Research Scientist at TikTok in Seattle, working on LLM / multimodal AI for e-commerce — SFT and reinforcement learning for multimodal LLMs, grounding and alignment strategies (verifiable rewards, retrieval-augmented reasoning, semantic consistency), and multimarket multimodal models for large-scale governance and content moderation.

I completed my Ph.D. in Electrical Engineering at NYU Tandon in May 2025, advised by Prof. Yao Wang in NYU Video Lab, and co-advised by Prof. Adeen Flinker on neural speech decoding. Before that I studied biology (XueTang program) and minored in statistics at Tsinghua University.

Research interests: multimodal large language models, LLM reasoning / alignment, neural speech decoding from ECoG/sEEG, and healthcare & brain–computer interface applications.


Resume


Contact

📍
Location
Bellevue, WA 98004
☎︎
Phone
347-889-0528
✉︎
💻
💼
LinkedIn

Work

TikTok — Seattle, WA Research Scientist, LLM / Multimodal AI for E-commerce | Jun 2025 – Present

  • Trained multimodal LLMs with SFT + RL (RLHF / reward modeling variants), focusing on reasoning quality and uncertainty calibration.
  • Researched grounding & alignment methods including verifiable reward design, retrieval-augmented reasoning (RAG), and semantic consistency regularization.
  • Built production-grade multimodal moderation systems for e-commerce, supporting multi-market deployment with improved precision/recall trade-offs and reduced labeling cost.

Education

New York University, Tandon School of Engineering — Brooklyn, NY Ph.D. in Electrical Engineering | Sep 2019 – May 2025

  • Advisor: Prof. Yao Wang (Video Lab)
  • Multimodal LLMs for medical imaging and LLM reasoning / alignment. Neural speech decoding with applications in healthcare and brain–computer interfaces.

Tsinghua University, School of Life Science — Beijing, China B.Sc. in Life Science (XueTang Program), Minor in Statistics | Sep 2014 – Jun 2019

  • XueTang program for top research-track students.

Publications

  1. Chen, X., Wang, R., Khalilian-Gourtani, A., Yu, L., Dugan, P., Friedman, D., Doyle, W., Devinsky, O., Wang, Y., Flinker, A. A Neural Speech Decoding Framework Leveraging Deep Learning and Speech Synthesis. Nature Machine Intelligence (2024).
  2. Wang, R., Chen, X., Khalilian-Gourtani, A., Yu, L., Dugan, P., Friedman, D., Doyle, W., Devinsky, O., Wang, Y., Flinker, A. Distributed feedforward and feedback cortical processing supports human speech production. PNAS (2023).
  3. Chen, X., Lai, Z., Ruan, K., Chen, S., Liu, J., Liu, Z. R-LLaVA: Improving Med-VQA Understanding through Visual Region of Interest. IJCNN (2025).
  4. Li, X., Chen, X., Fan, J., Jiang, E.H., Gao, M. ENCORE: Entropy-guided Reward Composition for Multi-head Safety Reward Models. AAAI (2026).
  5. Li, X., Yu, Z., Zhang, Z., Chen, X., Zhang, Z., Zhuang, Y., Sadagopan, N., Beniwal, A. When Thinking Fails: The Pitfalls of Reasoning for Instruction-Following in LLMs. NeurIPS (2026).
  6. Chen, X., Chen, J., Wang, R., Le, C., Khalilian-Gourtani, A., Jensen, E., Dugan, P., Doyle, W., Devinsky, O., Friedman, D., Flinker, A., Wang, Y. Transformer-based neural speech decoding from surface and depth electrode signals. Journal of Neural Engineering (2025).
  7. Wang, R., Chen, X., Khalilian-Gourtani, A., Chen, Z., Yu, L., Flinker, A., Wang, Y. Stimulus speech decoding from human cortex with generative adversarial network transfer learning. ISBI (Best Paper Finalist) (2020).
  8. Christou, P., Chen, S., Chen, X., Dube, P. Test Time Learning for Time Series Forecasting. NeurIPS 2024 Workshop on Time Series in the Age of Large Models (2024).
  9. Lin, Z., Wei, D., Jang, W.D., Zhou, S., Chen, X., Wang, X., Schalek, R., Berger, D., Matejek, B., Kamentsky, L., Peleg, A. Two stream active query suggestion for active learning in connectomics. ECCV (2020).
  10. Khalilian-Gourtani, A., Wang, R., Chen, X., Yu, L., Dugan, P., Friedman, D., Doyle, W., Devinsky, O., Wang, Y., Flinker, A. A corollary discharge circuit in human speech. PNAS (2024).
  11. Ni, H., Meng, S., Chen, X., Zhao, Z., Chen, A., Li, P., Zhang, S., Yin, Q., Wang, Y., Chan, Y. Harnessing Earnings Reports for Stock Predictions: A QLoRA-Enhanced LLM Approach. arXiv (2024).
  12. Meng, S., Chen, A., Wang, C., Zheng, M., Wu, F., Chen, X., Ni, H., Li, P. Enhancing Exchange Rate Forecasting with Explainable Deep Learning Models. arXiv (2024).
  13. Le, C., Gong, Z., Wang, C., Ni, H., Li, P., Chen, X. Instruction Tuning and CoT Prompting for Contextual Medical QA with LLMs. ICAHN (2025).
  14. Le, C., Zhao, Y., Emami, N., Yadav, K., Liu, T., Chen, X. VoxelFormer: Parameter-Efficient Multi-Subject Visual Decoding from fMRI. arXiv (2025).
  15. Emami, N., Khalilian-Gourtani, A., Qian, J., Ratouchniak, A., Chen, X., Wang, Y., Flinker, A. Machine Learning-Based Prediction of Speech Arrest During Direct Cortical Stimulation Mapping. arXiv (2025).
  16. Yang, K., Chen, X., He, J. Multimodal Fusion of Skeleton Dynamics and Clinical Gait Features for Video-Based Cerebral Palsy Severity Assessment. arXiv (2026).
  17. Meng, S., Chen, X. Artificial Intelligence and Systemic Risk: A Unified Model of Performative Prediction, Algorithmic Herding, and Cognitive Dependency in Financial Markets. arXiv (2026).
  18. Chen, X. Abundant Intelligence and Deficient Demand: A Macro-Financial Stress Test of Rapid AI Adoption. arXiv (2026).
  19. Chen, X., Meng, S. When AI Levels the Playing Field: Skill Homogenization, Asset Concentration, and Two Regimes of Inequality. arXiv (2026).
  20. Chen, Y., Chen, X., Wang, F., Jiao, N., Liu, J. VANGUARD: Vehicle-Anchored Ground Sample Distance Estimation for UAVs in GPS-Denied Environments. arXiv (2026).

See also Google Scholar.


Research and Work Experience

Large Language-and-Vision Assistant for BioMedicine with Vision Prompts — New York University

  • Developed a multimodal LLM for disease diagnosis from medical images, focusing on regions of interest (ROI). Pretrained on a large-scale medical image question-answering dataset, then instruction fine-tuned using ROI-labeled medical images and QA pairs.
  • Outperformed LLaVA-Med on four evaluation datasets in medical question answering and ROI localization with 3–8% accuracy and recall improvement.

Harnessing Earnings Reports for Stock Predictions with LLMs — New York University

  • Developed a QLoRA-enhanced LLM for stock predictions by integrating financial metrics, earnings transcripts, and other factors. Outperformed GPT-4 with a 19% accuracy increase and 7% weighted F1 improvement.
  • Implemented instruction-based fine-tuning for precise next-day stock movement predictions following earnings reports.

Radar and LiDAR Point Cloud Classification and Object Detection — NXP Semiconductors, Autonomous Driving Group | Summer 2022

  • Developed Radar and LiDAR point cloud classification and object detection models using PointNet++, PointPillar, and Point Transformer.
  • Implemented a semantic-information-guided sampling strategy; achieved 1–3% mAP improvement on KITTI and private datasets while reducing inference time by 20%.

Neural Speech Decoding and Synthesizing from sEEG/ECoG signals — Video Lab, New York University | 2020–2025 Supervisors: Yao Wang and Adeen Flinker · Grants: NSF IIS-2309057, NSF IIS-1912286

  • Developed differentiable and quantized speech re-synthesis frameworks for overt and imagined speech decoding, achieving 17–30% higher accuracy and fidelity than baselines.
  • Designed multi-subject neural decoding models with spatial–temporal attention for robust cross-participant generalization and improved intelligibility.
  • Proposed a disentangled contrastive learning approach to capture semantic and instance-level features from neural signals.

Efficient Instance Annotation for Connectomics — Lichtman Lab, Harvard University | Summer 2018 Supervisors: Hanspeter Pfister and Jeff Lichtman

  • Built a 3D U-Net for synapse detection on CREMI dataset — ranked 1st in CREMI contest.
  • Built 3D U-Net and 3D-CNN for synaptic connections and intracellular structures (e.g. mitochondria); constructed an active-learning annotation framework for proofreading.

Activities & Awards

  • 2025 — Dr. Li Annual Publication Award in Electrical EngineeringNew York University
  • 2024 — Dante Youla Award for Graduate Research Excellence in Electrical EngineeringNew York University
  • 2018–2022 — Teaching Assistant for Image and Video Processing, Medical Imaging, and Bioinformatics — NYU / Tsinghua
  • 2015–2019 — XueTang Scholarship ($10,000 for research) — Tsinghua University
  • 2017 — First Prize, The First National College Students’ Brain Computation and Application Competition — International
  • 2017 — First Prize, eMaize Challenge: Machine learning in genetic variation prediction — National
  • 2018 — Meritorious Winner, Mathematical Contest in Modeling (MCM) — International
  • 2016–2018 — Initiative Scientific Research Program ($8,000 for biomedical image analysis) — Tsinghua University

You may also find something interesting in my: