// AI × DATA × DELIVERY
把业务需求，变成跑得起来的 AI 系统。 Turning business needs into AI systems that actually run.
▸ ▌
我有 7 年数据运营与项目交付经验（PMP 认证），核心优势是把业务需求转化为可落地的解决方案：从数据采集、分析到自动化系统搭建，兼具软件实施与客户交付背景，能独立跑通「需求 → 方案 → 交付」全流程。 I have 7 years of experience in data operations and project delivery (PMP certified). My core strength is turning business needs into working solutions — from data collection and analysis to building automated systems — backed by a software implementation and client delivery background, running the full loop from requirement to shipped result.
我熟练运用大模型（Claude / GPT / DeepSeek）结合 Python 构建 AI 应用与自动化管线：行业情报采集分析、每日资讯聚合摘要、审计数据清洗、报表服务产品化……把 AI 能力转化为实际业务成果，而不止于「用工具提效」。 I build AI applications and automation pipelines with large models (Claude / GPT / DeepSeek) plus Python: industry intelligence collection and analysis, daily news digests, audit data cleaning, reporting productization — converting AI capability into business outcomes, not just "using AI tools to work faster."
需求 → AI + 代码 → 交付。这个闭环我独立跑通过很多次，你现在看到的这个网站，也是其中一件作品。 Requirement → AI + code → delivery. I've run this loop solo many times — the website you're reading is one of its products.
* 照片占位，之后替换 * Photo placeholder
2018.06 — 至今Present
主导公司数据支持全流程与经营分析：搭建覆盖人力、资产、产能、订单的数据监控体系，为近 200 人团队提供实时预警；引入大模型 + Python，把数据清洗、报表生成、分析报告撰写等重复流程自动化；负责部门级项目全流程跟进与跨部门协调；从 0 到 1 运营企业抖音官方账号（50+ 条作品，出镜主持与视频制作）。 Own end-to-end data operations and business analysis: built a monitoring system covering HR, assets, capacity and orders with real-time alerts for a ~200-person team; introduced LLMs + Python to automate data cleaning, report generation and analysis writing; run department-level projects with cross-functional coordination; grew the company's official Douyin account from zero (50+ videos as host, scriptwriter and editor).
2016.12 — 2018.05
早期「前向部署」式交付经验：直接对接客户、把产品落地到客户现场——软件定制与实施安装、SQL 数据导入、表结构设计与存储过程编写，以及客户培训与系统维护。 Early forward-deployed delivery experience — landing the product on-site with clients: software customization and implementation, SQL data imports, schema design and stored procedures, plus training and maintenance.
2015.12 — 2016.12
为政府部门及企业提供舆情监测服务：构建关键词体系、分析海量舆情数据，输出月度报告与突发事件专题报告。 Media monitoring for government and enterprise clients: keyword systems, large-scale sentiment analysis, monthly reports and incident-response briefs.
2012.09 — 2016.07
院学生会组织部副部长，负责学生事务文档管理及活动组织。 Deputy head of the student union's organization department — student affairs documentation and event organization.
独立完成「需求 → AI + 代码 → 交付」的完整闭环，用大模型 + 工程手段把信息处理自动化——而不止于「用 AI 工具提效」。 Each project runs the full loop solo — requirement → AI + code → delivery — automating real information work with LLMs and engineering, beyond just "using AI tools."
多源自动采集（公开 API + 动态页面渲染）、反爬应对、自动去重与增量入库，累计处理 1500+ 条行业数据；用大模型对非结构化文本做结构化抽取与分类，自动生成趋势分析报告。 Multi-source collection (public APIs + rendered pages) with anti-bot handling, dedup and incremental storage — 1500+ records processed. LLM-powered structured extraction and classification, with auto-generated trend reports.
定时自动化管线：每日聚合 AI 与金融领域多源资讯，大模型提炼摘要、生成结构化日报并自动推送，长期稳定运行；并延伸出「日报 → 快讯视频 / 播客」的内容自动化产线。 A scheduled pipeline aggregating AI and finance news daily — LLM summaries distilled into a structured briefing, pushed automatically and running reliably — extended into a briefing-to-video/podcast content line.
查看源码（视频/播客产线）View source (video/podcast pipeline) ↗把原本每周需专人耗时一天的数据导出与批量处理流程完全自动化：从需求对接到全流程跑通约 2 天完成交付，重复性工作从「每周固定一天」变为自动执行，并预留监察预警等扩展能力。 Fully automated a weekly routine that used to cost one person-day per week — delivered end-to-end in about two days from first requirement talk, with monitoring and alerting hooks reserved for extension.
运用大模型对审计所需数据进行清洗、核对与审核，自动生成报表与凭证（如全年逐月单据），在保证合规与准确的前提下显著提升处理效率。 LLM-assisted cleaning, cross-checking and review of audit data, auto-generating statements and vouchers (a full year of monthly documents) — much faster while staying compliant and accurate.
把本地自动化报告工具（每日自动推送数据报告）产品化上线：搭建在线报表、支持团队成员自助取数——完成从「本地脚本」到「线上自助数据服务」的演进，把一次性工具沉淀为团队可持续使用的服务。 Took a local auto-reporting tool online as a product: web-based reports with team self-service access — evolving a one-off script into a sustained data service the whole team relies on.
输入一篇论文，自动产出竖屏科普视频：Claude 写脚本、edge-tts 配音、Playwright 渲染，端到端无人工。这也是我第一个开源项目——代码可直接查看。 Feed in a paper, get a vertical explainer video — Claude scripts it, edge-tts voices it, Playwright renders it, fully automated. My first open-source project — code is public.
查看源码View source ↗你正在看的这个网站，从设计、编码到部署全程与 AI 结对完成——它既是作品集，也是一件展示 AI 协作开发的作品。 The site you're looking at was designed, coded and deployed in pair with AI — both a portfolio and a live demo of AI-collaborative development.
把模糊的业务需求拆成 AI 能执行的清晰任务。Break vague business needs into tasks AI can execute.
与 Claude 结对开发，把方案写成能跑起来的系统。Pair with Claude to turn the plan into a running system.
端到端验证真实效果，交付能持续运行的结果，而不是「代码写完了」。Verify real behavior end-to-end and ship something that keeps running — not just "code complete."
想聊聊合作或工作机会？欢迎随时联系。 Open to opportunities and collaborations — feel free to reach out.
karlyma102@gmail.com📍 广东中山📍 Zhongshan, Guangdong, China