LE

Leo

beijing

Joined April 2017

38 Followers

88 Following

AI Insights & Achievements

31/ 100

Coding for 9 years.

Profile README

Hi there 👋

AI Algorithm Engineer

Currently working at Zhongyun Weixun Technology Co., Ltd. as AI Algorithm Engineer

🎯 专注于 AI Agent平台、大模型工具链、推荐系统 和 大规模微服务架构

Focused on AI Agent Platform, LLM Toolchain, Recommendation Systems and Large-scale Microservices Architecture
💡 擅长构建从0到1的AI产品，主导过省文旅平台AI Agent产品技术开发

Expert in building AI products from scratch, led the development of provincial cultural tourism AI Agent platform
📊 参与过亿级流量推荐系统工程开发，经历APP日活从0到2300万的增长

Experienced in billion-scale recommendation system engineering, grew app DAU from 0 to 23M
📫 联系我: haiboz8@163.com

Contact me: haiboz8@163.com
💬 可以问我关于 AI Agent架构、推荐系统、大规模系统设计 的问题

Ask me about AI Agent Architecture, Recommendation Systems, Large-scale System Design

Connect with me

Core Skills

AI & 机器学习

AI & Machine Learning

熟悉多Agent系统、langchain/langgraph/llamaindex、mem0、langfuse、熟悉（Leader Agent / Sub Agent、Plan-Task / ReAct / Reflexion / TOT）等编排和推理模式

Proficient in multi-Agent systems, langchain/langgraph/llamaindex, mem0, langfuse, orchestration and reasoning patterns (Leader Agent / Sub Agent, Plan-Task / ReAct / Reflexion / TOT)
熟悉RAG检索召回（GraphRAG、BM25、Semantics Retrieval）

Experienced in RAG retrieval and recall (GraphRAG, BM25, Semantics Retrieval)
熟悉模型数据集构建、微调和评估、模型在线Serving（vllm、TriTon Serving）

Skilled in model dataset construction, fine-tuning, evaluation, and online serving (vLLM, Triton Serving)
熟悉大模型工具链（LiteLLM、reasoning mode、mcp/tools、context enginner）

Familiar with LLM toolchain (LiteLLM, reasoning mode, MCP/tools, context engineering)
熟悉推荐系统（多路召回、精排模型、多目标优化、在线推理、系统工程）

Expert in recommendation systems (multi-channel recall, ranking models, multi-objective optimization, online inference, system engineering)

后端开发

Backend Development

语言: Python, Java, Golang

Languages: Python, Java, Golang
架构: 微服务架构、大规模分布式系统

Architecture: Microservices architecture, large-scale distributed systems
基础设施: Kubernetes、CI/CD、DevOps

Infrastructure: Kubernetes, CI/CD, DevOps
数据: Hadoop、Spark、Flink、Redis

Data: Hadoop, Spark, Flink, Redis

前端开发

Frontend Development

Core Project Experience

Cultural Tourism AI Agent Platform

主导AI Agent平台技术架构和产品开发

Led AI Agent platform technical architecture and product development
多Agent编排服务、模型推理服务、RAG检索召回服务

Multi-Agent orchestration service, model inference service, RAG retrieval and recall service
Memory记忆管理、Tools/MCP服务

Memory management, Tools/MCP services
产品功能：语音智能问答、线路规划、数字人导游、AI拍照识景等

Product features: Voice Q&A, route planning, digital human tour guide, AI photo recognition

ShuaBao APP Recommendation System (Jumei Youpin)

从0到1搭建推荐系统，支撑千万级用户

Built recommendation system from scratch, supporting millions of users
用户人均观看时长提升100%（20分钟→55分钟）

Increased average viewing time by 100% (20min → 55min)
系统规模：500台召回服务器、100台精排服务器、200+ Redis实例

System scale: 500 recall servers, 100 ranking servers, 200+ Redis instances
数据规模：6PB，每日处理1TB数据

Data scale: 6PB, processing 1TB daily

Tech Stack

GitHub Stats

GitHub Stats

Top Langs

GitHub Streak

Featured Projects

workease cowork的开源版 Open source version of cowork	skill-hub skill导航站，收集了热门skill Agent skill navigation site, collecting popular skills	podgist 免费的热门的播客文字稿 Free transcripts of popular podcasts	deepresearch 开源的deepresearch产品 Open source DeepResearch product
vocalremover 在线的人声分离工具 Online vocal separation tool	ai-shop 通过AI生成商品图的工具站 AI-powered product image generation tool	rednote-ai-station 一站式小红书工作站，AI生成帖子和配图、多账号管理和发布 All-in-one Xiaohongshu workstation: AI post & image generation, multi-account management	deepsite 基于deepseek v3的网站生成器 Website generator based on DeepSeek V3