Xiaobing Chen

Xiaobing Chen

Research Scientist at Meta. I build systems that make AI collaborative, efficient, and resource-aware.

Bridging Massive-Scale ML & Autonomous Agents

I am a Research Scientist at Meta Platforms Inc. based in Menlo Park, CA, where I translate billion-scale user signals into measurable product wins. My work sits at the intersection of Massive-Scale Machine Learning and Autonomous Systems—bridging the gap between theoretical intelligence and practical, reliable application.

I believe the future of AI isn’t just about bigger models, but about smarter orchestration. Whether optimizing graph neural networks for billions of users or building autonomous agent fleets, I focus on creating systems that are robust, efficient, and impactful.

🚀 Impact at Scale (Meta)

At Meta, I architect and deploy machine learning systems that power core discovery and integrity experiences for billions of users. My focus is on turning complex signals into Tier-0 product wins.

  • Scale: Designed graph-based user representation learning systems that scale to billions of nodes, unlocking new capabilities for signal-sparse populations.
  • Efficiency: Architected offline evaluation frameworks and training pipelines that significantly reduced compute overhead while accelerating iteration velocity for the entire team.
  • Product: Led multiple high-impact launches across Instagram and Facebook, driving global engagement wins and integrity improvements through novel feature engineering and model architecture optimizations.

⚡ The Frontier: Agentic Systems

Beyond industrial ML, I am deeply invested in the future of Collaborative AI and Multi-Agent Systems. I believe the next leap in productivity comes from workflow consolidation—where specialized agents orchestrate complex tasks autonomously.

Current Exploration:

  • Autonomous Orchestration: Building upon “OpenClaw,” a personal exploration into multi-agent fleets that can plan, execute, and self-heal.
  • Workflow Consolidation: Moving beyond “chatbots” to “do-bots”—agents that integrate deep into tools and infrastructure to automate research and engineering loops.

🔬 Research Foundation

My engineering is grounded in rigorous academic research. I earned my Ph.D. from LSU, focusing on Federated Learning and Distributed Optimization.

DualGFL: Federated Learning with a Dual-Level Coalition-Auction Game
AAAI Conference on Artificial Intelligence (AAAI), 2025
The Insight: Solves decentralized training instability using game-theoretic auctions for fair device participation.

Enhancing Time Series Forecasting via Multi-Level Text Alignment with LLMs
DASFAA, 2025
The Insight: Bridges numerical sensor data with human-readable text to unlock LLM reasoning for complex forecasting.

Joint Device and Training Scheduling for Wireless Federated Learning
IEEE Internet of Things Journal, 2025
The Insight: A blueprint for edge AI deployment, balancing communication cost, energy, and performance.