Proof, not promises证据,不是承诺
Every project below shipped with a working demo, a verification report, and handover docs. Grouped by the three problems above. 下面每个项目都附带可运行 Demo、验证报告和交接文档。按上面三类问题归类。
Seven years of engineering across enterprise software, an AI product team, high-concurrency systems, and industrial robotics. I work at the seam where software meets hardware, and where messy business reality meets clean systems — and I prove the work with a running demo before you ever commit. 七年工程经验,横跨企业级软件、AI 产品团队、高并发系统、工业机器人。 我工作在两条交界线上——软件与硬件之间,以及混乱的业务现实与整洁的系统之间。 而且在你做任何承诺之前,我先用一个能跑的 Demo 把事情证明给你看。
Not a list of technologies — a list of business problems I take full ownership of. 这不是技术栈清单,而是我愿意完整扛下来的三类业务问题。
Devices, robots and IoT that speak engineer — coordinates, PLC, MQTT, upper-computer. I wrap them in interfaces (HMI) a non-technical operator can actually run, safely. 设备、机器人、IoT 说的是工程师的语言——坐标、PLC、MQTT、上位机。我给它们套上一层非技术人员也能安全操作的界面(HMI)。
"The expensive machine sits idle because nobody but an engineer can drive it.""那台贵机器一直闲着,因为除了工程师没人会开。"
Marketing, CRM, e-commerce, payment — platforms that each hold a piece and refuse to talk. I build the integrations, RPA and automations that make leads and orders flow instead of piling up. 营销、CRM、电商、支付——各握一块数据、互相不通气。我做集成、RPA 与自动化,让线索和订单流动起来,而不是堆积成山。
"Too many leads coming in, and we can't keep up with any of them.""客户越来越多,一个都跟不动。"
Not a chatbot toy. AI and automation aimed at the real leaks: missed calls, missed customers, and the repetitive software drudgery that eats your team's day. 不是聊天机器人玩具。把 AI 和自动化对准真正的漏洞:漏接的电话、流失的客户,以及吃掉团队一整天的重复性软件劳动。
"We're losing customers to missed calls and drowning in copy-paste work.""漏电话在丢客户,团队又淹没在复制粘贴里。"
Every project below shipped with a working demo, a verification report, and handover docs. Grouped by the three problems above. 下面每个项目都附带可运行 Demo、验证报告和交接文档。按上面三类问题归类。
The case studies above are the proof; here's the wider range. No price talk until we both know it's a fit. 上面的案例是证据;这里是更完整的范围。是否合适没确认之前,不谈价钱。
Touchscreen controls, fleet scheduling, Modbus / PLC — machines ordinary staff can run safely.触屏控制、车队调度、Modbus / PLC——让普通员工也能安全操作机器。
Event-driven microservices, database migrations, tuning from 50 to thousands of QPS.事件驱动微服务、数据库迁移、把吞吐从 50 调到数千 QPS。
Voice or documents → structured data, with a confidence score that flags what a human should check.语音或文档 → 结构化数据,用置信度评分标出需人工复核的部分。
Give Claude or GPT real tools — pause a campaign, query a system — with a mock mode for safe demos.给 Claude / GPT 装上真工具——暂停广告、查询系统——自带可安全演示的 Mock 模式。
Does what the API can't: multi-step UI flows, auto-2FA, headless / cloud runners, no password held.做 API 做不到的:多步 UI 流程、自动过 2FA、无头 / 云端运行,不保存密码。
REST, webhooks, POS / CRM / payment — the tools you already use, finally talking to each other.REST、webhook、POS / CRM / 支付——让你已在用的工具终于互通。
n8n, scheduled scripts, alerting — repetitive work runs itself and tells you when something breaks.n8n、定时脚本、告警——重复工作自动跑,出问题会主动通知你。
Messy pages → clean JSON / CSV, on a schedule, resilient to layout changes.杂乱页面 → 干净的 JSON / CSV,可定时,能扛住页面改版。
Not sure yours fits? Tell me the problem and what it's costing you — I'll give you an honest read on whether I can solve it, and a demo usually follows. 不确定你的需求算不算?告诉我问题、以及它正让你付出什么代价——我会诚实判断能不能解,通常接着就是一个 demo。
Trusting a remote engineer is hard. I solve it with process, not promises. 信任一个远程工程师很难。我用流程解决它,而不是用承诺。
You see working software before you spend a dollar on implementation. Mock modes let me prove the mechanics without touching your credentials. 在你为开发投入一分钱之前,先看到能跑的软件。Mock 模式让我不碰你的任何凭据就能证明机制可行。
No radio silence. I work async and send short Loom updates showing exactly what's built — bugs and risks surfaced early, not hidden. 绝不失联。我异步工作,定期发简短 Loom 录屏展示实际进展——bug 和风险提前摊开,而不是藏着掖着。
I've reverse-engineered MySQL Binlogs to recover 100% of lost production data during an emergency. When something goes wrong, I'm the person who stays and fixes it. 我曾在应急中逆向 MySQL Binlog、100% 恢复丢失的生产数据。真出了问题,我是那个留下来把它修好的人。
Handover docs cover how to run it, how to migrate it to your own accounts, and every credential to replace. No lock-in, no dependency on me. 交接文档写清如何运行、如何迁移到你自己的账号、以及需替换的每一个凭据。零锁定,不依赖我。
I'm a full-time independent engineer — this is my craft and my livelihood, not a side gig. I take end-to-end ownership of a problem and I don't disappear when it gets hard. My commitment is to you, directly. 我是全职独立工程师——这是我的手艺、也是我赖以为生的事业,不是副业。我对一个问题端到端负责,越难越不会消失。我的承诺,是直接对你的。
A display-panel maker with ~15% of China's market · HQ IT, where I built several of the company's internal ERP systems.一家占中国显示器市场约 15% 的面板厂 · 总部 IT,我在这里搭了公司内部的多套 ERP 系统。
I re-architected a fragile monolithic ERP/WMS into event-driven microservices, sitting with the business teams from spec to launch — downtime failures down ~90%, throughput 50 → 5,000+ QPS. Talking every day to the people who actually used the system is how I learned to build what a business needs, not just what a ticket says.我把一套脆弱的单体 ERP/WMS 重构成事件驱动的微服务,从需求到上线全程和业务方坐在一起——宕机故障降约 90%,吞吐 50 → 5,000+ QPS。每天和真正用系统的人对话,就是我练出"做业务要的、而不只是工单写的"这项能力的方式。
An AI product company · years before ChatGPT, building field-extraction on Baidu's open-source model. One of the engineers on it.一家 AI 产品公司 · 远早于 ChatGPT,基于百度开源模型做要素抽取。我是其中一名工程师。
The raw model extracted at ~60% accuracy. We pushed parameters to their limit to reach ~80%, then layered in real business-rule understanding to hit ~95% — industry-leading for its time. Not comparable to today's LLMs, but the value was real: work that took 20 people entering data became one person reviewing. That's where I learned that engineering rigor plus domain understanding beats a raw model — and the standards of real product collaboration. (Still on good terms with its founders.)原始模型抽取准确率约 60%。我们把参数调优到极限做到约 80%,再叠加对业务规则的真正理解,做到约 95%——当时业内一流。和今天的大模型不能比,但价值是真的:原来 20 个人的录入工作,变成 1 个人的审核。就是在这里,我懂了"工程严谨 + 业务理解,胜过原始模型",也吃透了真实产品协作的规范。(和创始团队至今关系很好。)
A fintech company that raised $100M each from Goldman Sachs and Tencent · its core product matched investors, founders and roadshows. One of 8 backend engineers.一家分别拿过高盛、腾讯各一亿投资的金融科技公司 · 核心产品撮合投资人、创业者与路演。我是 8 名后端工程师之一。
I led production database migrations across seven microservices and owned high-concurrency paid consumer flows. Delivering large, non-standard work reliably — in phases, milestone by milestone, without dropping the ball — is the discipline I built here.我主导了 7 个微服务的生产库迁移,负责高并发的付费 ToC 核心链路。把庞大、非标的工作可靠地交付——分阶段、一个里程碑一个里程碑地推进、不掉链子——这项本事是在这里练出来的。
An industrial AMR maker whose AGV/arm shipments reached global top-5 in 2025 · I was the sole developer holding every internal system.一家 2025 年 AGV/机械臂出货量进入全球前五的工业 AMR 厂商 · 我是 hold 住其全部内部系统的唯一开发。
I worked at the seam between software and physical machines and cleaned up years of legacy until the internal systems ran without a babysitter — including recovering 100% of lost production data by reverse-engineering MySQL Binlogs. My work there is backed by a formal recommendation letter from the CTO — available on request.我工作在软件与物理机器的交界处,把积压多年的烂摊子理顺,直到内部系统不再需要人盯着——包括逆向 MySQL Binlog、100% 恢复丢失的生产数据。这段工作有该公司 CTO 的正式推荐信背书——可应请求提供。
Betting my craft on my own name. All in — this is what I do now.把手艺押在自己的名字上。全力以赴——这就是我现在做的事。
My permanent home base — every project and relationship centers here, not on any single platform.我永久的大本营——每个项目和客户关系都以这里为中心,不依赖任何单一平台。
Productizing my voice-to-data pipeline, a storefront for packaged services, and direct client work — all anchored here.把语音转数据管道产品化、开一个打包服务的店铺、直接和客户合作——全部以本站为锚。
Here's exactly what working with me looks like — the terms, and how to start. 和我合作具体长什么样——条款,以及怎么开始。
