曹曦文
Xiwen Cao Xiwen Cao
曹曦文

Product Manager · UX Designer · Full-Stack Developer

"以共情驱动设计，以数据验证决策，
实现以用户为中心的产品设计。"

"Empathy-driven design, data-driven decision-making,
and user-centered product design."

密西根大学安娜堡分校敏捷开发与用户体验设计硕士，上海交通大学电子与计算机工程学士。具备从需求洞察到产品落地的全链路思维，希望成为技术与用户之间的桥梁。

M.S. in User-Centered Agile Development at the University of Michigan, Ann Arbor; B.S. in Electronic & Computer Engineering from Shanghai Jiao Tong University. End-to-end product thinking from insight to launch, blending technical depth with user experience.

曹曦文 Xiwen Cao

Background

教育背景 & 技能

Education & Skills

密歇根大学安娜堡分校

University of Michigan, Ann Arbor

敏捷开发与用户体验设计 · 硕士

M.S. Agile Development & UX Design

2024.08 – 2026.05

GPA 4.0 / 4.0

上海交通大学

Shanghai Jiao Tong University

电子与计算机工程 · 学士 + 数据科学 · 辅修

B.S. Electronic & Computer Engineering · Minor in Data Science

2021.09 – 2025.08

GPA 3.5 / 4.0 · 校级三好学生 · 本科生优秀奖学金

GPA 3.5 / 4.0 · Merit Student · Academic Scholarship

技能图谱

Skills

编程语言

Languages

PythonC++SQLJavaScriptHTML/CSSC#

设计工具

Design & Tools

FigmaUnityPhotoshopPremiereTableauCanva

语言

Spoken

中文 · 母语英语 · TOEFL 110
Chinese · NativeEnglish · TOEFL 110

Experience

实习经历

Internships

美团

Meituan

聚合配送产品 · 产品运营 + 产品经理

Integrated Delivery · Product Ops & Product Manager Intern

2025.05 – 2025.08

帮助中心功能优化 费力度 ↓6% · 工单 ↓800/月

针对帮助中心内容匮乏、结构简单导致用户学习成本高的问题，通过 5+ 竞品调研定义高优需求，推动一期优化方案落地。预计上线后产品使用费力度 降低 6%，功能类客服工单 减少 800 单/月。

方案设计分为两个层次：前端侧重信息架构重组与反馈机制搭建；后端新增运营配置后台为本次方案的核心亮点，由本人在需求分析过程中主动识别问题并提出。

核心亮点 · 运营配置后台

优化前 · 内容更新链路
  • 运营发现问题
  • 提交更新需求
  • 前端实现（延迟较长）
  • 内容上线

✗ 每次更新依赖前端资源

优化后 · 运营自助配置
  • 运营发现问题
  • 登录后台自助编辑内容
  • 内容即时上线

✓ 无需开发介入，随时更新

方案推进过程：技术团队初期以研发资源受限为由拒绝纳入该功能。经进一步评估，提出两项论据推动决策——① 一期帮助文档数量约为现有的 2–3 倍，且预期随产品迭代持续增长；缺乏配置后台将导致每次内容变更均需占用开发资源，长期累计成本远高于一次性建设投入；② 调研发现平台现有「站内信发送后台」与所需功能高度重合，底层均依赖同一富文本编辑平台，存在显著的能力复用空间，可大幅压缩研发工作量。最终方案获得技术团队认可并纳入一期排期。

预期结果

6%
整体费力度降低
40%×14% 推导
800
功能类工单减少 / 月

ℹ 以上均为上线前预测值。费力度降幅基于调研数据量化推导，工单减少量基于历史功能类工单分析估算，待功能上线后持续追踪实际数据。

鲜花订单处理提效 处理效率 +50%+

基于真实跟单体验与商家深度沟通，设计"条形码扫码直达"及"待处理订单 Tag"机制，大促期间订单处理效率 提升 50%+，有效提升商家高峰期履约稳定性。

由于鲜花品类要求，商家需要进行实拍返图（入口在订单详情页面），原有订单处理流程链路较长。因此设计了条形码扫码直达功能，大大提高了订单处理效率。

同时，增加“待处理”Tag，展示骑手已到店订单，帮助商家快速定位订单，减缓大促期间门店骑手拥堵压力。

系统性需求调研 走访 5+ 家商家

主导竞品核心功能 MLB 对标分析，深入走访 5+ 家 品类代表商家，包含酒水、日料、菜市场等多元化品类，最终输出结构化洞察报告，为产品迭代提供市场对标与真实场景双维度依据。

Help center optimization −6% effort · ~800 tickets/mo · tap heading for details

Addressed thin content and flat IA that raised learning cost by prioritizing needs from 5+ competitor reviews and shipping a first-phase redesign. Expected: −6% perceived effort and ~800 fewer functional support tickets per month.

The plan had two layers: customer-facing IA refresh and coverage; a new ops CMS was the main innovation—I flagged the publish bottleneck in discovery and proposed it.

Highlight · Ops admin

Before · Content pipeline
  • Ops finds a gap
  • File eng request → wait for slot
  • Build (weeks of delay)
  • Ship content

✗ Every update needed engineering

After · Self-serve config
  • Ops finds a gap
  • Edit in admin
  • Publish immediately

✓ No eng for routine updates

How we got it shipped: Engineering first declined for capacity. Two arguments shifted the decision: (1) phase-one docs were ~2–3× current volume and would keep growing—without a CMS, repeated eng work would cost more than building once; (2) an existing in-app message admin shared the same rich-text stack, so reuse could cut scope. The CMS was then approved for phase one.

Expected impact

6%
Lower perceived effort
Derived 40%×14%
800
Fewer feature tickets / mo.
0
Eng cycles for copy updates

ℹ Pre-launch estimates only. Effort math and ticket deltas were aligned with biz and data; we will track actuals after launch.

Flower order handling efficiency +50%+ efficiency · tap heading for details

From ride-alongs and merchant interviews, designed barcode deep-links and a “pending order” tag flow, raising handling efficiency by 50%+ and stabilizing peak-time fulfillment.

Traced merchant steps from accept to dispatch, identified scanning and status recognition as bottlenecks, and documented low-fi flows plus edge cases for engineering sizing. Ran walkthroughs with pilot merchants and refined tag rules and default filters before rollout.

Structured discovery 6 merchant visits

Led MLB-style benchmarking of competitor core flows and field visits to 6 representative merchants, delivering a structured insights report for roadmap and real-world context.

电通中国

Dentsu China

技术与数据实习生

Technology & Data Intern

2024.04 – 2024.06

数据管理模块设计 检查耗时 ↓近 80% · 点击标题展开详情

独立设计集多源数据实时可视化于一体的工单数据看板，每日工单数据检查耗时 缩短近 80%，提升内部管理效率。

使用SQL查询并提取重要指标，基于公司内部数据分析平台进行数据可视化。与使用者迭代看板布局，确保关键指标固定在首屏，减少切换成本。

数据平台迁移 迁移策划与文档

参与内部数据平台迁移的策划，包括前端页面优化、数据整理和功能映射设计，并撰写操作手册及文档。

Data management module ~80% faster checks · tap heading for details

Designed a multi-source, near real-time dashboard for ~100 staff, cutting daily data-check time by ~80%.

Cataloged sources and refresh cadence per line of business, aligned metric definitions and alert thresholds, and layered access controls. Iterated layouts with users so daily-critical KPIs stay above the fold.

Platform migration Planning & docs

Contributed to an internal analytics migration: UI refresh, data cleanup, feature mapping, and user documentation.

Projects

项目经历

Projects

产品经理 · SI612 课程项目

Product Owner · SI612 course project

FLOWMATE

针对业余钢琴练习者在翻页与犯错时面临的专注中断与情绪挫败痛点，主导用户调研与产品定义，通过问卷与访谈识别三大核心痛点（翻页打断演奏、错误记录繁琐、缺乏情绪支撑），提炼设计方向并确定功能优先级；同时使用 Vibe Coding 独立实现可交互 Demo 原型。

Identified three core pain points faced by amateur pianists — focus disruption from manual page turning, the burden of tracking mistakes, and lack of emotional support during struggle — through surveys and user interviews. Led product definition and feature prioritization based on research insights; independently built an interactive demo prototype using Vibe Coding.

用户访谈User interviews用户旅程Journey Map需求定义Product DefinitionVibe Coding

核心组员 · 毕业设计银奖

Core member · Capstone silver award

基于大模型与 AR 的人机协同拆解系统

LLM & AR human–machine collaborative disassembly

构建集成 LLM、数字孪生与 AR 的人机协同拆解系统，解决电池拆解安全性与效率问题。负责多模态模型评估、Web 端交互界面搭建及基于 DiffCSE+Cross-Encoder 的双阶段语义验证机制设计。

Built a collaborative disassembly system combining LLMs, digital twin, and AR for safer, faster battery teardown. Owned multimodal model evaluation, web UX, and a two-stage semantic verification pipeline (DiffCSE + cross-encoder).

Gemini 2.0Prompt EngineeringYOLOv11DiffCSE

产品经理 · 创新比赛

Product Lead · Innovation competition

HEART GARDEN

针对年轻女性分手后的情感痛点，主导设计融合情感陪伴与自我成长的移动应用。通过 10+ 用户访谈构建用户画像，确定"任务种子""社区匿名伙伴匹配""AI 情感支持"三大核心功能，围绕手绘风"分手花园"概念完成 UI 与交互设计。

Led a mobile app blending emotional support and self-growth for young women after breakups. From 10+ interviews, defined three pillars — “task seeds,” anonymous peer matching, and AI emotional support — and shipped hand-drawn “breakup garden” UI/UX.

NLPBFS 算法BFSRAGFigma

Get in Touch

联系方式

Contact

欢迎随时联系我。

Feel free to reach out anytime.

xiwencao@umich.edu (+86) 135-1522-8556 微信: xixiw_wechat WeChat: xixiw_wechat