Chaoran Jiang
EN

Data Science × AI Products × Markets

Chaoran Jiang

I build products with AI as leverage, and read markets with a poker player's discipline: no edge, no bet. 我用 AI 当杠杆做产品,用牌手的纪律看市场:无优势,不下注。

UW-Madison Data Science '27UW-Madison 数据科学 2027 届 Buy-side intern, Shanghai '262026 上海买方投研实习 EN / 中文
Live demo · my product solving a semester实时演示 · 我的产品在排课 0 conflicts ✓0 冲突 ✓

UWM Class Selection Assistant. Upload a degree audit, get conflict-free schedules that satisfy graduation requirements, in 0.8s. UWM 一键选课助手。上传毕业审核报告,0.8 秒排出零冲突、满足毕业要求的课表。

0s
full-semester solve整学期排课耗时
0
automated checks green项自动化检查全绿
0
AI agents, one review个 AI agent 并行审查
0
national patents项国家专利
01 · The Product产品

Not a demo. A shipping product, built solo, with an agent fleet. 不是 demo,是一个人带着 agent 舰队交付的真产品。

The schedule you watched assemble above is my constraint-solver engine working on real catalog data. It plans what students spend days agonizing over, and it never returns an empty answer. 开屏那张自己拼好的课表,就是我的约束求解引擎在真实课程数据上运行。学生纠结几天的选课,它一秒排完,而且永远不会甩给你一个空结果。

4,395
Constraint Engine约束求解引擎

OR-Tools CP-SAT over a normalized catalog, 4,395 courses linked to grade history. When sections fill up it degrades honestly instead of failing. OR-Tools CP-SAT 跑在规范化课程库上,4,395 门课打通历史成绩数据。选不上时诚实降级,绝不摆烂报错。

×32
Built with Agentsagent 造出来的

A 32-agent adversarial review attacked the engine before launch: 20 confirmed bugs found and fixed. The build holds 826 checks green. 上线前用 32 个 agent 对抗审查围攻引擎,确认并修掉 20 个 bug;构建保持 826 项检查全绿。

EN·中
Full Stack, Bilingual全栈 · 双语

FastAPI + PostgreSQL behind, Next.js in front, EN/中文 UI throughout. Same hands that designed the product wrote the SQL. 后端 FastAPI + PostgreSQL,前端 Next.js,全站中英双语。产品设计和 SQL 出自同一双手。

Python · FastAPI · PostgreSQL · OR-Tools CP-SAT · Next.js · TypeScript · Claude Code

02 · The Method方法

One person, operating like a team. 一个人,跑出一支团队的产能。

AI is my leverage, not my autopilot. I design the system, agents do the labor, adversarial review keeps everyone honest. AI 是杠杆,不是自动驾驶。我设计系统,agent 出苦力,对抗审查让所有产出过得了质检。

1 TASK finder · logic finder · data finder · perf skeptic ×3 skeptic ×3 … ×32 agents 20 FIXED
32 agents · 51 findings→ 20 confirmed & fixed

AIOS, my personal AI operating systemAIOS:我的个人 AI 操作系统

Daily briefs, email triage, project memory across everything I do. It distills my corrections into new operating rules every week, human-reviewed. 日报、邮件分诊、全项目记忆图谱。每周把我的纠正蒸馏成新规则,人工审核后生效,系统自己长大。

Adversarial by default默认对抗验证

Every important claim gets skeptic agents hired to refute it. What survives, ships. What doesn't, never reaches the user. 重要结论一律雇「杠精 agent」专职反驳。驳不倒的才上线,驳倒的永远到不了用户面前。

Ship, not demo交付成品,不交付演示

Checks, honest degradation, bilingual UI, real data. The bar is "a stranger can use it," not "it worked once on my machine." 检查全绿、诚实降级、双语界面、真实数据。标准是「陌生人能用」,不是「在我电脑上跑通过一次」。

03 · The Markets市场

Research with an invalidation line drawn first. 先画认错线,再做研究。

Quant strategy · paper-traded量化策略 · 模拟盘 +34% · max DD -8%
+40% +20% 0 invalidation line · no thesis, no position认错线 · 无 thesis 不建仓
NO EDGE, NO BET无优势 · 不下注

AI-assisted quant pipelineAI 辅助量化管线

Daily US-equity selection from a 60-ticker universe into 3-5 positions. Interpretable features only, LLMs translate constraints into explicit rules, freshness checks block look-ahead bias. 60 只标的池里每日选出 3-5 只建仓。只用可解释特征,LLM 把投资约束翻译成显式规则,数据新鲜度检查掐死未来函数。

AI application-layer thesisAI 应用层投资 thesis

One question: whose budgets does AI eat, and who gets re-rated? Contested names (Adobe, Unity, Duolingo), validated on middle indicators, never on price. 只问一个问题:AI 在吃谁的预算、谁会被重估?盯争议股(Adobe、Unity、Duolingo),用中间指标验证,不看股价自我说服。

Sizing markets for a buy-side team给买方团队做市场测算

Built the team's AI-video market model in Shanghai: global top-down, explicit penetration assumptions, every number traceable to a primary source. 在上海给投研团队搭 AI 视频市场测算:全球口径自上而下,渗透率假设全部显式,每个数都能溯源到一手材料。

04 · The Story经历

Founder at 15. Analyst at 19. Builder since. 15 岁创业,19 岁做分析师,然后一直在造东西。

2026

Investment Research Intern投研实习生

Buy-side asset manager · Shanghai买方资管机构 · 上海

Market sizing for AI video generation: clean front report, fully sourced back workbook. Weekly investment-logic reviews with a senior mentor. CFA Level I in preparation. AI 视频生成行业测算:前台报告干净,后台底稿全溯源。每周和资深 mentor 过投资逻辑。CFA 一级备考中。

2024

Data Analyst Intern数据分析实习生

xMEMS Labs · Santa Clara, CA

Cleaned and transformed finance datasets in Python and loaded them into the accounting system; contributed to the team's Tableau operations dashboard; ran SEO keyword analysis in SEMrush. Python 清洗财务数据并录入记账系统;参与搭建团队的 Tableau 运营看板;做 SEMrush SEO 关键词分析。

2020

Founder, Ignite 3D Hologram创始人 · Ignite 3D 全息

Hangzhou, China杭州

Designed a 3D hologram display, secured two national patents, and pitched the business plan to two venture investors. 设计 3D 全息显示装置,拿下两项国家专利,向两位创投投资人做了商业计划路演。

2023 - 2027

University of Wisconsin-Madison

B.A. Data Science, expected June 2027数据科学本科,2027 年 6 月毕业

Coursework in data science modeling, data science programming, linear algebra, and micro- and macroeconomics. 课程包括数据科学建模、数据科学编程、线性代数与微观/宏观经济学。

05 · Beyond Work工作之外

Flip a card. 翻张牌看看。

Poker
AA

Where the discipline comes from.
Only play hands with an edge.
纪律的来源。
只打自己有优势的牌局。

Vocal
AA

Six years of pop & R&B.
Top 4, university competition.
六年流行与 R&B。
校级歌唱比赛前四。

Fitness
AA

Five years, 5×/week.
Tracked like any other dataset.
五年,每周五练。
把身体当数据集追踪。

Court & Stage
AA

Badminton & dance.
Reflexes and stage presence.
羽毛球与舞蹈。
练反应,也练台风。

If this beat the last twenty resumes you read, let's talk. 如果这比你今天看的前二十份简历有意思,聊聊。

Email copied邮箱已复制