HUMAN · IN · THE · LOOP

THE SHORT VERSION · SKIP THE WORLD

Weirui Kuang

邝炜瑞

Researcher & Senior Machine Learning Engineer

Alibaba Tongyi Lab阿里巴巴通义实验室

I build agent harnesses for long-horizon tasks: steerable loops, human-in/on-the-loop systems, and multi-agent orchestration.

MULTI-AGENTLLMFEDERATED LEARNINGGRAPH LEARNING

CURRENTLY

Building better agent harnesses, with agents.

Long-horizon loops. Human-in/on-the-loop systems. Agents orchestrated to work well together.

Scholar ↗GitHub ↗

ROOM 01 / OPEN SOURCE

Open source,
made for real work.

Projects are ranked by live GitHub stars. They began as research or engineering problems and grew into tools people can actually use.

RANK 01↗

Personal AI Assistant

QwenPaw

A self-hostable personal AI assistant for multiple chat apps with extensible capabilities.

RANK 02↗

Agent Framework

AgentScope

A production-ready framework for building agents you can see, understand and trust.

RANK 03↗

Production Runtime

AgentScope-Runtime

Secure tool sandboxing, Agent-as-a-Service APIs, scalable deployment and observability.

ROOM 02 / RESEARCH

Old paths,
open questions.

Google Scholar ↗

Multi-Agent Systems

Harnesses, orchestration, and agents that can work together without getting in one another’s way.

Large Language Models

Fine-tuning, evaluation, and agent systems that hold up outside a demo.

Federated Learning

Training across data silos without moving private data into one place.

Graph Learning

Learning useful representations from large, connected data.

SELECTED PUBLICATIONSKDD · ICML · NeurIPS · VLDB · TKDE

01KDD 20241ST AUTHOR

FederatedScope-LLM: A Comprehensive Package for Fine-tuning Large Language Models in Federated Learning

02TKDE 2024CO-FIRST AUTHOR

When Transformer Meets Large Graphs: An Expressive and Efficient Two-View Architecture

03ICML 2023CO-FIRST AUTHOR

FedHPO-Bench: A Benchmark Suite for Federated Hyperparameter Optimization

04VLDB 20236TH AUTHOR

FederatedScope: A Flexible Federated Learning Platform for Heterogeneity

05NeurIPS 20223RD AUTHOR

pFL-Bench: A Comprehensive Benchmark for Personalized Federated Learning

06KDD 2022CO-FIRST AUTHORBEST PAPER

FederatedScope-GNN: Towards a Unified, Comprehensive and Efficient Package for Federated Graph Learning

ROOM 03 / TIMELINE

The road
that got me here.

Tongyi Lab, Alibaba Group · Hangzhou, China

EXPERIENCE

2021 — NOW

Alibaba Tongyi Lab

Senior Machine Learning Engineer

Multi-agent systems, federated learning for LLMs, AutoML and graph representation learning.

2021

Alibaba DAMO Academy

Research Intern

Graph representation learning and KDD Cup 2021 OGB-LSC.

2020 — 2021

Alibaba Cloud

Research Intern

Block storage cardinality estimation.

EDUCATION

2018 — 2021

Renmin University of China

M.E. in Computer Science and Technology

Graph learning; advised by Zhewei Wei.

2014 — 2018

Renmin University of China

B.S. in Mathematics and Applied Mathematics

School of Information.

ROOM 04 / BLOG

Notes, mistakes,
and things worth keeping.

Field notes from building agents, doing research, and shipping software. More are on the way.

FIELD NOTES01READ THE NOTES ↗