Shanghai Jiao Tong University
BSc Finance (Minor: Information Management)
Core Courses
Business Statistics, Database & Big Data, Data Collection & Visualization, Computational Text Analysis, Algorithm Analysis & Design, Operations Research, AI Fundamentals, Econometrics, Fixed Income Securities, Securities Investment Analysis, Venture Capital & Financial Innovation, Financial Engineering, Corporate Finance
EDHEC Business School
Master in International Finance & Management (Dual Degree)
GPA 15.66/20 (Top 15%) | Outstanding Chinese Alumni Scholarship (50% tuition)
Core Courses
Python Portfolio Analysis, Fintech, Financial Modeling, VBA, M&A Analysis, Financial Analysis, Climate Change Science & Economics
Natixis (Natixis Corporate & Investment Banking)
AI Risk Analytics Intern
- Metrics framework & instrumentation: Decomposed KYC/compliance processes into funnel stages (submission -> acceptance -> review -> supplementary -> approval/rejection), defining throughput, SLA, backlog, and anomaly rate metrics to support "identification -> strategy adjustment -> review" cycles.
- Automated data pipeline (3,000+ lines of code): Multi-source data cleaning, validation, and scheduled refresh, supporting daily reports and anomaly tracking, reducing ~20% manual operations.
- Observability dashboard (Power BI): Designed drill-down paths by region/product/channel/reason code with threshold alerting, enabling R&D, operations, and compliance teams to collaborate around a unified fact table.
- AI workflow monitoring loop: Established "metric definition -> anomaly detection -> root cause analysis -> retrospective review" workflow, strengthening productized understanding of rule monitoring, data feedback mechanisms, and process optimization strategies.
GE Vernova
Finance Digital Audit Intern
- Report productization: Excel modeling + VBA automation for forecast/operations reports (version control, metric validation, anomaly alerts), reducing errors and improving delivery stability.
- Process governance & controlled rollout: Mapped data integrity, authorization approval, and audit trail risk points, producing remediation checklists and execution steps (corresponding to payment link canary/rollback/audit tracking).
- Modular analysis: Decomposed variance analysis into reusable templates (driver decomposition, sensitivity, scenario comparison), improving growth/commercialization analysis efficiency.
Allinpay Payment Fintech
AI Risk Product Manager & Business Analyst Rotation
- Product iteration: Assisted in upgrading YunShangTong (2021 Shanghai Fintech Innovation Regulatory Sandbox Project), driving "business-finance integration" digital transformation, coordinating business and financial requirements, planning transformation roadmap and driving system API development, completing core module alignment (30% of total tasks), improving cross-department response efficiency by over 35%.
- Trained and fine-tuned financial risk control models using knowledge graphs and Graph Neural Networks (GNN), combining MetaPath, community detection, and GNN algorithms to uncover complex risk relationships, deployed in anti-fraud, AML, and intermediary fraud detection scenarios.
- Report writing: Independently prepared and presented 12 industry briefings to the Board, covering policy analysis, trend tracking, and industry forecasts, plus 5 special reports (cross-border payments, digital finance, etc.), providing data-driven strategic recommendations to senior management.
- Regulatory draft research: Managed card organization relations (UnionPay), participated in cross-border payment and digital RMB research, led drafting of "QR Code Payment Interim Regulations" based on policy analysis and international case studies, providing research evidence for institutional development.
Caizhen Securities Co., Ltd.
Risk Control Intern
- Assisted in financial statement risk identification and compliance review, investigating potential risk areas around revenue recognition, inventory valuation, and cash flow to support project risk assessment.
- Compiled risk identification and assessment reports, summarizing key risk issues and improvement recommendations to support internal reporting and follow-up remediation.
AML Sentinel
AML Intelligent Monitoring Workbench
Designed a 5-view workbench for bank compliance teams, covering the full alert-to-STR pipeline. Integrated IBM AML real data (84 alerts + 71 cases) with ECharts network graph visualization.
- RAG pipeline on Anthropic Claude API: BM25 + CJK bigram tokenization over 55 regulatory chunks from 4 sources; SSE streaming + tool_use structured generation for CAMLMAC-format STR auto-drafting
- PBOC threshold rules engine (cash CNY 50K / transfer CNY 2M), 4-tier risk scoring, ECharts force-directed network graph with fan-out / fan-in / cycle typology detection
- Stack: Next.js 14 / TypeScript / Tailwind CSS / Anthropic Claude API (RAG + tool_use) / SSE Streaming / ECharts / Vercel
ComplianceIQ
Compliance Research Knowledge Workbench
Product suite with AML Sentinel: AML Sentinel finds risks, ComplianceIQ finds rules. Designed evidence layering separating legal provisions from reference materials; 25 bilingual enforcement cases with cross-module navigation.
- Two-layer retrieval (BM25 regulation ranking + server-side case recall) + claude-sonnet-4-6 SSE streaming; enforced 5-section structured answers (conclusion / legal basis / cases / action items / disclaimer)
- Evidence layering: legal/regulatory provisions (AML Law, PBOC Rules) strictly separated from reference materials (KPMG, FATF); cross-product deep linking with AML Sentinel
- Stack: Next.js 14 / TypeScript / Claude API / SSE / BM25 + CJK Bigram / i18n bilingual / Vercel
AI Agent Werewolf Championship
Ranked 24th / 164 teams
AI Agent competition: designed strategy templates with state machines (memory / hypothesis / action) and established a strategy-update / offline-eval / online-verify iteration loop.
- Built role-specific strategy engines outputting confidence-scored decisions with reasoning, reducing crash risk under adversarial conditions
- A/B tested prompt/strategy variants for stability vs reward tradeoffs; applied gradual rollout methodology from production AI systems
Multi-Agent Research System
LangChain/LangGraph Agent Orchestration
Researcher-Analyst-Writer three-role Agent collaboration pipeline with independent system prompts, toolsets, and memory modules orchestrated via directed graph state machine.
- Agent toolchain: Web Search, PDF parsing, data analysis; supports breakpoint resumption and intermediate result caching
- Stack: Python / LangChain / LangGraph / Claude API / Next.js frontend
Financial Document RAG Q&A
Chinese Regulatory Q&A System
Regulatory Q&A system for fintech compliance teams with PBOC / SAFE / NPC multi-document retrieval.
- Hybrid BM25 + vector retrieval (0.3:0.7) supporting cross-lingual Chinese-English queries
- Structured JSON output (answer / sources / confidence) with inline citation cards for traceability
Risk & Finance
Data & Tech Tools
Languages
EDHEC Campus Ambassador & Chairman, Chinese Society
Raised €2,000 in sponsorship and organized the Spring Festival Gala with 200+ attendees. Led outreach campaigns for prospective Chinese students and served as official EDHEC Campus Ambassador.