作品

Work

国家二级博物馆的官网,三套自研系统。都在运行中,不是概念稿。鼠标放到图上可播放几秒的动态展示。

A national grade-two museum's website and three in-house systems. All running, none of them concepts. Hover an image for a few seconds of motion.

  • 东莞市福木源紫檀博物馆官网首屏:千年紫檀,国家二级博物馆 Home screen of the Fumuyuan museum site

    东莞市福木源紫檀博物馆 · fumuyuan.中国客户项目Client work

    国家二级博物馆的官方网站,从零构建至上线

    The official website of a national grade-two museum, from zero to launch

    东莞市福木源紫檀博物馆(国家文物局认定的国家二级博物馆,2018 年落成开馆,总建筑面积四万平方米)官方网站:藏品典藏、文创雅集、研习体验、馆藏简介与参观预约,中英双语。Next.js 全栈实现,16 个路由,已上线运行。

    The official site of the Dongguan Fumuyuan Red Sandalwood Museum (a national grade-two museum accredited by China's National Cultural Heritage Administration; opened in 2018, 40,000 m² of floor area): collection, cultural products, study visits, museum profile and visit booking, in Chinese and English. Full-stack Next.js, 16 routes, live.

    已上线 · fumuyuan.中国 · 16 个路由

    Live · fumuyuan.中国 · 16 routes

    访问站点 Visit the site
  • grounded-chat自研In-house

    缺乏依据的语句,由系统自行标注

    Every shaky sentence gets flagged

    面向企业内部资料的问答系统。本地向量检索,回答附带引用;每一句再与检索到的段落逐一比对,缺乏依据的语句即时标注。本站右下角的「问 Kailoo」运行的是同一条链路。

    A knowledge assistant for internal documents. Local retrieval, answers with citations, and every sentence is scored against the retrieved passages so the unsupported ones are flagged. The 'Ask Kailoo' in the corner of this site runs the same chain.

    223 项自动化测试通过,另有 9 项依赖 Postgres 环境。

    223 automated tests passing, plus 9 that need Postgres.

    了解更多 Learn more
  • gig-radar自研In-house

    线索自动生成,人工只审核最后一步

    Leads surface on their own, a person reviews only the last step

    从多个来源采集线索,由模型评分,按渠道起草,发出后自动扫描回复。每日无人值守运行,每一步均有记录。

    Pulls leads from several sources, scores them, drafts per channel, then scans the replies. Runs daily unattended, and every step is logged.

    61 项自动化测试全部通过。

    61 automated tests, all passing.

    了解更多 Learn more
  • AI 科普视频生产线AI explainer video line自研In-house

    一集 AI 科普视频,从脚本到成片全链自动

    An AI explainer episode, script to final cut, end to end

    AI 科普视频生产线:口播脚本按段落切分镜,版式化图形(bento 卡片、大数字、时间线、对比)由渲染引擎生成,配音与字幕自动对齐;每一镜先经机器审帧,标记的镜才交人工复核。

    The AI explainer video line: the narration script is cut into shots, templated graphics (bento cards, big numbers, timelines, comparisons) are rendered by the engine, voice-over and subtitles are aligned automatically; every shot passes a machine frame review first and only flagged shots go to a person.

    course-0001:16 段脚本 · 21 镜 · 逐镜机器审帧 · 成片 3:41

    course-0001: 16 script lines · 21 shots · machine frame review on every shot · 3:41 final

    对应服务 The service behind it

告诉我们哪一个环节最耗时

Tell us which step costs you the most

联系我们Contact us