Building reliable AI agents, practical LLM applications, and verifiable workflows.
🎓 Wuhan University of Technology ('28) · 📍 Wuhan, China · 🎯 Seeking AI Agent / LLM App Internships
- 🤖 Focus: Building reliable AI agents, tool-using workflows, RAG systems, and developer productivity tools.
- ⚙️ Engineering: Plan-and-Execute orchestration, structured outputs (
Pydantic), failure handling, and benchmark evaluations. - 🌟 Open Source: 10 merged PRs across 7 upstream projects — n8n, DeepTutor, rea, PixelRAG and other agent / developer-tooling ecosystems.
Python (asyncio / Pydantic / pytest) · Agent Workflows & Tool Calling · Hybrid RAG & RRF · CI/CD & Mocking
| Project | Description | Tech Stack |
|---|---|---|
| CountyResearchAI | LLM Research Agent: Plan-and-Execute runtime, dynamic question-tree planning, multi-search provider routing, and 260+ regression tests. | Python Agent Runtime pytest |
| courselm | Document Automation Pipeline: Slides parsing → LLM distillation → Pydantic validation, featuring per-stage caching, resume-on-break, and Windows CI. | Python Pydantic CI/CD |
Other Projects: MicroGlow-Translate (ESP32 TinyML, Provincial Grant) · esp-screenpilot (ESP32-S3 HID) · student-os
Issue-linked fixes, regression tests included — 10 merged PRs in 7 repositories.
- n8n-io / n8n (Workflow Automation Platform · 209k★)
- Fixed error payload loss and
TypeErrorin the Splunk node under "Continue On Fail", with HTTP regression tests (#40368 · Merged).
- Fixed error payload loss and
- morluto / rea (Agentic Reverse Engineering · 432k★)
- Reclassified malformed residual-unknown graphs from
unreadable_outputtoinvalid_input, surfacing the offending Evidence ID and the correct recovery path (#6321 · Merged).
- Reclassified malformed residual-unknown graphs from
- HKUDS / DeepTutor (Lifelong AI Tutoring · 55k★)
Also merged in PixelRAG #171 / #173 · codex-with-chatgpt #434 · AntigravityManager #335 · antigravity-skin #1




