Peng Cheng / AI Agent 工程师AI Agent Engineer
专注储能 BMS、工业运维与企业自动化，让大模型从对话框走进生产系统。 Battery systems, industrial operations and enterprise automation — moving LLMs out of the chat box and into production.
01 — 精选作品Selected work
从需求文档直接生成 BMS 嵌入式代码的 AI Agent 框架。 An agent framework that turns BMS requirement documents into embedded C.
储能电站运维闭环：告警 → 诊断 → 知识检索 → 工单。 A closed O&M loop for storage plants: alarm → diagnose → retrieve → ticket.
测试 SOP 智能生成系统：照片分组、AI 建议与 Word 自动排版。 Test-SOP generation: photo grouping, AI review and automated Word layout.
AI 驱动的 A 股量化分析平台。 An AI-driven quant analysis platform for China A-shares.
更多项目More projects
企业 AI 学习、资讯、实践与治理中台，内置页面感知助手。Enterprise hub for AI learning, news and governance, with a page-aware assistant.
↗ReAct 范式邮件 Agent：分类、回复草稿、信息抽取与提醒。A ReAct email agent: triage, reply drafts, extraction and reminders.
↗AI 内容流水线：拆解故事 DNA，驱动原创、剧本与 AI 视频。AI content pipeline: story DNA to original writing, scripts and AI video.
↗02 — 关于About
我关注的问题是：如何让大模型在严肃的工业场景里可靠地干活。储能告警诊断、BMS 参数适配、测试文档生产，这些场景对准确性、可追溯性和部署环境都有硬约束。所以我更在意 Agent 的工程化：清晰的阶段编排、规则与模型的分工、人在回路的确认，以及能长期运维的完整系统。
The question I care about: how do you make LLMs do reliable work in serious industrial settings? Alarm diagnosis, BMS parameter adaptation, test documentation — all carry hard constraints on accuracy, traceability and deployment. So I focus on agent engineering: explicit stage orchestration, a clear split between rules and models, human-in-the-loop checks, and systems built to be operated for years.
BMS 故障诊断、健康评分、运维 Agent 与嵌入式代码生成。嵌入式背景（STM32 / CAN）让链路一直延伸到设备侧。BMS diagnosis, health scoring, O&M agents and embedded code generation — with an STM32 / CAN background that reaches down to the device.
ReAct 与工具调用、AutoGen 多智能体编排、RAG 知识库、LLM-first 文档解析。ReAct and tool calling, AutoGen multi-agent orchestration, RAG, LLM-first document parsing.
FastAPI / tRPC 后端、Vue / React 前端，PostgreSQL、MongoDB 与 Docker 私有化部署。FastAPI / tRPC back ends, Vue / React front ends, PostgreSQL, MongoDB and Docker on-prem.
03 — 联系Contact
AI Agent 落地、储能智能化，或者只是聊聊技术，都欢迎。Agents in production, smarter energy storage, or just a good technical chat.