Master's student at University of Pennsylvania (MCIT, May 2026) with a background in Quantitative Statistics from Peking University. I build backend systems, distributed pipelines, and AI-powered applications.

📍 Philadelphia, PA  |  📧 nicole.lili731@gmail.com  |  LinkedIn  |  🏆 Top 3 @ MCIT Hackathon 2025

Languages Python Java JavaScript/TypeScript SQL R C/C++ Go

Frameworks & Systems React Next.js Node.js Express.js FastAPI Spring Boot Prisma ORM Kafka Redis RabbitMQ Elasticsearch Tailwind CSS

Database & Infrastructure PostgreSQL MySQL MongoDB HBase AWS GCP Azure Docker Kubernetes CI/CD Linux

Python · PyTorch · Gemma 3n · FastAPI · OpenCV · GCP  |  Google DeepMind Hackathon

Multi-agent retail monitoring system using a fine-tuned Gemma 3n vision-language model to process live camera feeds — distinguishes staff from customers in real time, calculates staffing shortfalls, and delivers alerts to a live dashboard.

• Dual-agent architecture: FloorAlertAgent for frame analysis + InsightAgent for business logic
• End-to-end pipeline: OpenCV ingestion → FastAPI backend → polling dashboard

TypeScript · Next.js 15 · Express.js · Prisma ORM · PostgreSQL · Docker

Full-stack marketplace connecting sponsors and publishers, built on a PNPM monorepo. Features role-based access control, automated DB migrations, and Docker containerization for consistent environments.

• React 19 frontend + Express.js backend in a unified monorepo
• Secure auth with Better Auth, distinct dashboards per user role

JavaScript · React · Node.js · REST APIs · PostgreSQL · MongoDB

Data platform for exploring Philadelphia rental markets. Combines structured and document data, integrating crime statistics, 311 service requests, and rental prices for neighborhood-level insights.

• Hybrid architecture: PostgreSQL for listings/stats + MongoDB for user profiles
• Real-time filtering and neighborhood ranking via optimized backend APIs