I'm a Java backend engineer who has gradually moved from traditional enterprise systems into enterprise AI engineering.

Today, most of my work is around turning LLM capabilities into systems that can actually be used inside a company: AI assistants, Agents, RAG, knowledge bases, workflow orchestration, MCP/tool integration, model gateways, observability, permissions and production delivery.

I care less about AI demos and more about one question:

How do we make AI useful, reliable and maintainable in real business workflows?

My current focus includes:

• 🧠 AI Coding — Claude Code, Codex, QwenPaw, DSH, Pi, ChatGPT and coding agents as part of my daily engineering workflow
• 🤖 AI Agents — ReAct, tool calling, MCP, multi-step task execution and agent orchestration
• 📚 RAG & Knowledge Systems — retrieval, reranking, vector databases, enterprise knowledge access and permission-aware retrieval
• 🧩 Dify / LLM Application Platforms — workflow design, model integration, production deployment and platform engineering
• ☕ Java Backend Engineering — Spring Boot, WebFlux, Dubbo, MyBatis, distributed systems and service integration
• 🔎 Search & Data Infrastructure — Elasticsearch / OpenSearch, Milvus, MySQL, Oracle and Redis
• 🐳 Infrastructure & Delivery — Docker, Nginx, Linux, gateways, SSE / streaming and production troubleshooting

I believe the next generation of enterprise software will not simply "add a chatbot".

It will combine:

LLM + Knowledge + Tools + Workflow + Permissions + Observability + Human Collaboration

into systems that can understand intent, retrieve context, call business capabilities and complete real tasks.

That's the direction I'm exploring every day.

AI-first workflow, engineering-driven delivery.

Tool Calling · Prompt Engineering · Workflow Orchestration · Memory · Planning · Multi-step Reasoning

Embedding Models · Reranking Models · Vector Search · Hybrid Search · OpenAI-compatible APIs · SSE / Streaming

Java · Spring Boot · WebFlux · Dubbo · MyBatis-Plus · MySQL · Oracle · Redis · Docker · Nginx · Linux

• chatgpt-proxy-launcher — macOS launcher / network helper for ChatGPT Desktop connectivity scenarios
• claude-pulse — lightweight status panel for monitoring Claude Code sessions
• ResponseRewriter — Chrome extension for intercepting and rewriting API responses
• langchain4j — exploration and learning around Java + LLM application development
• dify — tracking and exploring the Dify ecosystem

• MiniSpringMVC — learning Spring MVC internals by building a simplified implementation
• spring-source-learn — Spring internals and source-code learning
• dubbo-learn — Dubbo learning and experiments
• springboot-learn-integration — Spring Boot integration experiments
• tech-sharing-mq — messaging / MQ related technical exploration

• posy-pip-picture-book — an interactive children's picture-book experiment
• weather-for-kids — a kid-friendly weather project
• futaosmile.github.io — personal site / technical notes

• Turning AI from demo → product → platform → infrastructure
• Building systems that solve actual problems instead of chasing buzzwords
• Understanding how things work under the hood
• Using AI coding tools to increase engineering leverage, not replace engineering judgment
• Continuous learning, writing, experimenting and shipping