·
Builders & Technical Leaders
Xingyao Wang
Co-founder & CAIO |All Hands AI (OpenHands)
Built OpenHands (69K+ stars), the leading open-source AI software development agent. Invented CodeAct (ICML 2024) for code-based agent actions.
5papers
GitHubxingyaowwResearch Papers
5 papersarXiv
Executable Code Actions Elicit Better LLM Agents
2402.01030·2024-02-01
arXiv
OpenHands: An Open Platform for AI Software Developers as Generalist Agents
2407.16741·2024-07-23
arXiv
MINT: Evaluating LLMs in Multi-turn Interaction with Tools and Language Feedback
2309.10691·2023-09-19
arXiv
Training Software Engineering Agents and Verifiers with SWE-Gym
2412.21139·2024-12-30
arXiv
The OpenHands Software Agent SDK: A Composable and Extensible Foundation for Production Agents
2511.03690·2025-11-06
Biography
Deep Research
Xingyao Wang (born 1999, Fujian, China) is the Co-founder and Chief AI Officer of All Hands AI, building OpenHands -- the leading open-source AI software development agent with 69,000+ GitHub stars and 300+ contributors. He is a PhD candidate in Computer Science at the University of Illinois Urbana-Champaign, advised by Prof. Heng Ji and collaborating with Prof. Hao Peng. He earned his B.S. in Computer Science and Data Science from the University of Michigan, where he graduated with University Honors and received an Honorable Mention for the CRA Outstanding Undergraduate Researcher Award. His research focuses on AI agents that interact with computers through code, assist humans, and continuously self-improve. He is the first author of the foundational CodeAct paper (ICML 2024), which proposed using executable Python code as a unified action space for LLM agents -- the architectural insight that underpins OpenHands. He co-founded All Hands AI in 2024 alongside Robert Brennan (CEO) and Graham Neubig (CMU professor), raising $5M seed led by Menlo Ventures / Anthology Fund (Anthropic) and $18.8M Series A led by Madrona. Previously interned at Google Research, Microsoft, ByteDance, and Tencent.
AI Software Engineering AgentsCode-Based Agent ActionsOpen Source AI InfrastructureLLM Agent BenchmarksMulti-Turn Tool InteractionSandboxed Agent ExecutionSWE-Bench EvaluationAgent Self-ImprovementMultimodal LLM Pre-TrainingReasoning Models
Timeline
16 Research16 total
2026
2025
2025-04Research
OpenHands paper presented at ICLR 2025 as a conference paper2025-07Research
OpenHands agent surpassed the last human developer contributing code to its own repository2025-07Research
Published SWE-Gym: Training Software Engineering Agents and Verifiers (accepted ICML 2025)2024
2024-01Research
MINT and CRAFT papers accepted to ICLR 2024 -- evaluating LLMs in multi-turn interaction with tools2024-03Research
OpenDevin (later renamed OpenHands) launched on GitHub as open-source AI software development agent2023
Key Contributions
OpenHands
The leading open-source AI software development agent (69K+ GitHub stars, 300+ contributors). Autonomously solves software engineering tasks end-to-end -- writing code, running commands, and browsing the web in sandboxed environments. First open AI agent to surpass 50% on SWE-Bench Verified.
CodeAct
ICML 2024 Oral paper proposing executable Python code as a unified action space for LLM agents, achieving up to 20% higher success rates than JSON/text alternatives. The foundational architecture behind OpenHands.
MINT Benchmark
ICLR 2024 benchmark for evaluating LLMs in multi-turn interaction with tools and language feedback, providing a rigorous framework for assessing agent capabilities across iterative problem-solving.
SWE-Gym
Training framework for software engineering agents and verifiers (ICML 2025). Provides environments for training and evaluating AI agents on real-world software engineering tasks.
OpenHands Software Agent SDK
A composable and extensible SDK for building production-grade AI agents, providing the building blocks for multi-agent orchestration and tool integration.
CRAFT
ICLR 2024 paper on customizing LLMs by creating and retrieving from specialized toolsets, enabling dynamic tool creation for novel tasks.
Notable Quotes
“
Instead of AI demanding constant attention, I prefer autonomous AI that solves problems independently.
“
When it comes to dev tools, open source generally wins.
“
We're building open source AI agents that can tackle all the toil involved in an engineer's day-to-day work, so developers can focus on what they do best -- creative problem-solving.
“
Xingyao's deep empathy for the developer creates the perfect storm to tackle this market.
12 sources(click to expand)
Xingyao Wang -- personal websitexingyaoww (Xingyao Wang) -- GitHub profileOpenHands: An Open Platform for AI Software Developers (ICLR 2025)Executable Code Actions Elicit Better LLM Agents (ICML 2024)All Hands AI $5M Seed AnnouncementOpenHands Raises $18.8M Series A -- The SaaS NewsMadrona Investment in OpenHandsZ Potentials -- Xingyao Wang: 99-born PhD Founder in AI ProgrammingOne Year of OpenHands -- All Hands AI BlogXingyao Wang -- Google ScholarMINT: Evaluating LLMs in Multi-turn Interaction (ICLR 2024)SWE-Gym: Training Software Engineering Agents (ICML 2025)
Research generated March 19, 2026
Builders & Technical Leaders/Xingyao Wang
All Profiles