毕羽西Fashion Systems Archive

服装设计 / 品牌系统 / AI 原生产品研究

从服装设计走向Fashion AI 研究与产品实践

起点

SCAD、巴黎与米兰的服装设计专业教育培训背景

实践

5 年高级服装设计与品牌打造经验

延伸

在同济大学把时尚知识转向 Fashion AI 研究

01SCAD / 巴黎 / 米兰

025 年高级时尚设计实践

03StePO-Rec 第一作者 · EI 收录

04IQON-3000 Recall@1 +28.3%

Proof

代表性成果

集中呈现教育背景、秀场经历、论文发表与研究贡献。

学业查看简历

SCAD - ECSCP - 同济

SCAD 奖学金 · SCAD GPA 3.54/4 · 同济 GPA 4.91/5

秀场查看背景

NYFW Supima 2014

SCAD 全校唯一作品入选 Supima Fashion Show

论文查看论文

StePO-Rec

个性化穿搭推荐方向论文 · 第一作者 · ACM MM · CCF-A · EI 收录

综述查看论文

RAG Survey

行业首篇综述论文 · 高被引 RAG 综述论文 · 核心方法图示贡献

Selected Cases

核心 Fashion AI 项目

Research → System → Product → Playful Branch

StePO-Rec

个性化穿搭推荐方法设计研究

知识驱动推荐

第一作者 · ACM MM · Recall@1 +28.3% · MAP +32.8%

StePO-Rec preview

StePO-Rec

个性化穿搭推荐方法设计研究

第一作者 · ACM MM · Recall@1 +28.3% · MAP +32.8%

研究基础知识驱动推荐

FashionAgent

基于个性化穿搭推荐场景的多智能体系统原型

多智能体拆解

多角色拆解 · 可解释推荐 · OUTFITTED 方法与架构基础

FashionAgent preview

FashionAgent

基于个性化穿搭推荐场景的多智能体系统原型

多角色拆解 · 可解释推荐 · OUTFITTED 方法与架构基础

系统原型多智能体拆解

OUTFITTED

AI Agent 时尚产品设计与产品化收敛

可用产品收敛

AI产品独立开发 · 从需求理解到偏好记忆的多 Agent 产品链路

OUTFITTED preview

OUTFITTED

AI Agent 时尚产品设计与产品化收敛

AI产品独立开发 · 从需求理解到偏好记忆的多 Agent 产品链路

产品收敛可用产品收敛

包治百病

送礼推荐决策应用/ playful branch

vibe coding 应用实验

独立开发 · 风格测试、品牌匹配与像素风界面语言

包治百病 preview

包治百病

送礼推荐决策应用/ playful branch

独立开发 · 风格测试、品牌匹配与像素风界面语言

趣味分支vibe coding 应用实验

Research Visuals

复杂研究结构的视觉可视化

Taxonomy → Timeline → Framework → Evaluation → Collaboration

把复杂研究，压缩成可被快速理解的视觉信息

内容提炼：从复杂论文、综述与方法体系中抓出真正值得保留的核心结构

研究逻辑：把时间线、范式差异、方法关系与推理路径组织成可浏览的系统

视觉表达：让图示既专业清晰，又具备编辑感与审美完成度

01

Fragments

z_Modular RAG-3 — • Modular RAG-3

z_Modular RAG-总结图 — • Modular RAG-总结图

Rainbow Potion 7-needle — • Rainbow Potion

flare — • Flare

02

Paradigms

Comparison between the three paradigms of RAG — • RAG Paradigms

Hierarchical multi-agent execution — • Hierarchical Multi-Agent

Dual fine-tuning — • Dual Fine-Tuning

retrieverFT — • Retriever FT

Decision-support workflow — • Decision-Support Workflow

U-NIAH framework — • U-NIAH

ITER-RETGEN — • Iter-RetGen

03

Taxonomy

Human-AI collaboration evolution — • Human-AI Evolution

Task taxonomy of LLMs in urban planning — • Urban Planning Taxonomy

Case of current Modular RAG — • Modular Case

z_Modular RAG-1 — • Modular RAG-1

Cost quadrant diagram — • Cost quadrant diagram

Linear RAG flow pattern — • Linear RAG Flow

z_RAG_CASE图-英文-0103修改 — • RAG_CASE图-英文-0103修改

04

Reasoning

Advantages of Combining RAG with Reasoning — • RAG + Reasoning Gains

toc — • TOC

Practical Guide on RAG & Reasoning — • Practical Guide

Different spatial scales in urban planning tasks — • Spatial Scales

From LLM to RAG and then to RAG + Reasoning — • From LLM to RAG Then to RAG + Reasoning

Implementation and optimization of the synergy — • Synergy Optimization

z_Modular RAG-框架图 — • Modular RAG-框架图

z_databricks — • Databricks

05

Evaluation

loop — • loop

Representative co-creative frameworks — • Co-creative Frameworks

The four evaluation dimensions — • four evaluation dimensions

z_Modular RAG-2 — • Modular RAG-2

RAG vs other model optimization methods — • RAG vs Optimization

z_OPENAI — • OpenAI

Comparison between three RAG paradigms — • Three RAG Paradigms

Typical conversational assistant architectures — • Assistant Architectures

06

Architectures

The purpose of the synergy — • purpose of the synergy

Post-retrieval branching flow pattern — • Post-Retrieval Branching

generatorFT — • generatorFT

Patterns of Synergy between RAG and Reasoning — • Synergy Patterns

Timeline of studies on RAG-reasoning synergy — • RAG-Reasoning Timeline

The RAG flow in REPLUG — • RAG flow in REPLUG

z_graphRAG — • graphRAG

The typical pipeline of multi-modal fusion — • typical pipeline of multi-m…

07

Bias / Data

z_Modular RAG-4 — • Modular RAG-4

Generative de-biased matching framework — • Bias Reassignment

z_RAG Milestone 0821 — • RAG Milestone 0821

Data collection and user submission process — • Data Submission Flow

A timeline of existing RAG research — • timeline of existing RAG re…

conditional flow pattern — • Conditional Flow

08

Branches

RRR — • RRR

Pre-retrieval branching flow pattern — • Pre-Retrieval Branching

selfRAG — • SelfRAG

comprehensive and practical settings — • Practical Settings

Cases of Naive RAG and Advanced RAG — • Naive vs Adv.

01

Fragments

z_Modular RAG-3 — • Modular RAG-3

z_Modular RAG-总结图 — • Modular RAG-总结图

Rainbow Potion 7-needle — • Rainbow Potion

flare — • Flare

02

Paradigms

Comparison between the three paradigms of RAG — • RAG Paradigms

Hierarchical multi-agent execution — • Hierarchical Multi-Agent

Dual fine-tuning — • Dual Fine-Tuning

retrieverFT — • Retriever FT

Decision-support workflow — • Decision-Support Workflow

U-NIAH framework — • U-NIAH

ITER-RETGEN — • Iter-RetGen

03

Taxonomy

Human-AI collaboration evolution — • Human-AI Evolution

Task taxonomy of LLMs in urban planning — • Urban Planning Taxonomy

Case of current Modular RAG — • Modular Case

z_Modular RAG-1 — • Modular RAG-1

Cost quadrant diagram — • Cost quadrant diagram

Linear RAG flow pattern — • Linear RAG Flow

z_RAG_CASE图-英文-0103修改 — • RAG_CASE图-英文-0103修改

04

Reasoning

Advantages of Combining RAG with Reasoning — • RAG + Reasoning Gains

toc — • TOC

Practical Guide on RAG & Reasoning — • Practical Guide

Different spatial scales in urban planning tasks — • Spatial Scales

From LLM to RAG and then to RAG + Reasoning — • From LLM to RAG Then to RAG + Reasoning

Implementation and optimization of the synergy — • Synergy Optimization

z_Modular RAG-框架图 — • Modular RAG-框架图

z_databricks — • Databricks

05

Evaluation

loop — • loop

Representative co-creative frameworks — • Co-creative Frameworks

The four evaluation dimensions — • four evaluation dimensions

z_Modular RAG-2 — • Modular RAG-2

RAG vs other model optimization methods — • RAG vs Optimization

z_OPENAI — • OpenAI

Comparison between three RAG paradigms — • Three RAG Paradigms

Typical conversational assistant architectures — • Assistant Architectures

06

Architectures

The purpose of the synergy — • purpose of the synergy

Post-retrieval branching flow pattern — • Post-Retrieval Branching

generatorFT — • generatorFT

Patterns of Synergy between RAG and Reasoning — • Synergy Patterns

Timeline of studies on RAG-reasoning synergy — • RAG-Reasoning Timeline

The RAG flow in REPLUG — • RAG flow in REPLUG

z_graphRAG — • graphRAG

The typical pipeline of multi-modal fusion — • typical pipeline of multi-m…

07

Bias / Data

z_Modular RAG-4 — • Modular RAG-4

Generative de-biased matching framework — • Bias Reassignment

z_RAG Milestone 0821 — • RAG Milestone 0821

Data collection and user submission process — • Data Submission Flow

A timeline of existing RAG research — • timeline of existing RAG re…

conditional flow pattern — • Conditional Flow

08

Branches

RRR — • RRR

Pre-retrieval branching flow pattern — • Pre-Retrieval Branching

selfRAG — • SelfRAG

comprehensive and practical settings — • Practical Settings

Cases of Naive RAG and Advanced RAG — • Naive vs Adv.

Trajectory

从服装形式，走向系统形式

服装设计 → 品牌 / 系统思维 → AI 研究 → AI 原生产品原型

我的训练起点是时尚设计，但我真正长期关注的是结构、关系与系统如何生成风格。过去几年，这种关注逐渐延伸到研究方法、知识建模与 AI 产品原型实践。

01

设计基础

服装设计训练塑造了我对廓形、材质、色彩与受众感知的敏锐判断。

02

行业实践

奢侈与礼服工作强化了我对品牌语境、产品审美与系统化表达的理解。

03

研究转向

研究生阶段，我开始把这些直觉翻译为知识工程、检索、推理与可解释推荐方法。

04

当前整合

现在，我把这些跨界积累继续转化为 AI 原生工具与界面，让它既懂领域，也真正可用。