Leon Nie

About

Leon Nie (Leon Nie)
AI engineer, founder, and software engineering student at Tsinghua University

I study software engineering at Tsinghua University and will continue into master's study from 2026 to 2028. I co-founded Cavelight, where I lead engineering work on AI education products and enterprise AI systems.

Education

  • Bachelor of Software Engineering, Tsinghua University, 2022-2026, GPA 3.7/4.0.
  • Master's study confirmed for 2026-2028.
  • Recipient of the Science and Technology Innovation Excellence Scholarship.

Experience

  1. Cavelight

    Co-founder & Chief Technology Officer

    • Built and led an 8-member engineering team, setting up R&D processes, CI/CD, GitLab, and production infrastructure from scratch.
    • Architected and deployed an AI Teacher Assistant platform with a Rust/Vue stack, piloted in several schools in Beijing and deployed at Qian Xuesen Middle School.
    • Led delivery of enterprise AI systems including a smart risk-control project for CNOOC, covering data governance, monitoring, alerts, and workflow closure.
    • Designed multi-agent and RAG systems for compliance automation, policy consultation, and intelligent monitoring.

  2. Geoming AI

    AI Engineer Intern

    • Built a medical AI agent for pediatric consultation, assisted diagnosis, and prescription workflows.
    • Used LangChain, Qdrant, RAG, SGLang, and vLLM to improve retrieval quality and inference performance.
    • Developed benchmark tooling for evaluating AI models on pediatric medical diagnosis tasks.
    • Implemented a GraphRAG-based patient knowledge graph to support personalized consultation.

  3. Alchemy Pay

    AI Engineer Intern

    • Developed KOL CryptoPulse for crawling, analyzing, and maintaining crypto KOL tweet data.
    • Integrated LLMs for tweet classification, coin recognition, market sentiment analysis, and macro context retrieval.
    • Built CryptoCopilot, an AI-driven crypto quantitative analysis platform for Q&A, strategy backtesting, and real-time market analysis.
    • Implemented backend APIs and CLI tooling in Golang.

  4. Tsinghua University IoTDB Group

    Research Intern

    • Worked on tree-to-relational model compatibility in time-series databases.
    • Helped build an MCP server for LLM-IoTDB interaction with Python and Java.
    • Participated in designing an NLP query interface for industrial IoT data.

Technical

  • Agent applications, RAG systems, GraphRAG, context engineering, LangChain, AutoGen, Qdrant, and Faiss.
  • LLM deployment and inference with SGLang, vLLM, Ollama, local model serving, fine-tuning, and speculative decoding.
  • Python, C/C++, Golang, Java, C#, Dart, TypeScript, Rust/Vue, Kubernetes, Nacos, Flask, and production deployment.

Contact

nla22@mails.tsinghua.edu.cn
GitHub
Cavelight