Li (William) Zhu

Data Science @ Penn State | Building AI-Powered Products (Python, React, Claude API)

State College, Pennsylvania, United States

About

Experience

  • Full-Stack Developer at IdeaLab
    Mar 2026 - Present · 4 mos

    Built a full-stack simulation engine that tests product ideas against 30 AI-driven consumer personas across 8 behavioral archetypes, producing structured adoption and market insight reports. Developed a Python/FastAPI backend with a tick-based evaluation pipeline combining deterministic scoring models with bounded LLM qualitative reasoning Integrated Claude API (Haiku for NPC reactions, Sonnet for reports); optimized cost to under $0.15/simulation through request batching and model tiering Built a React/TypeScript frontend with real-time SSE streaming, interactive force-directed social graph, and in-character NPC chat Designed a variant comparison system for A/B testing product positioning with side-by-side metrics and AI-generated explanations

  • Software Developer at Aegis Trader
    Jan 2026 - Present · 6 mos

    Built an automated cryptocurrency trading bot in Python with live Coinbase API integration (REST + WebSocket), executing real-time buy/sell orders across BTC and altcoin markets. Implemented multi-strategy execution engine with EMA crossover and trend-pullback strategies, auto-selecting based on live market regime and volatility Designed a risk management framework with 9 pre-trade gates including position limits, fee efficiency checks, loss-streak pauses, and daily loss caps Built a real-time web dashboard with SSE streaming, performance charts, trade history, and an AI copilot Wrote 38 unit tests covering strategy logic, risk gates, and order execution

  • Founder & Lead Software Developer at Self-employed
    May 2024 - Nov 2025 · 1 yr 7 mos

    Designed and built a DIY AI robot focused on affordability and modularity, targeting consumer-grade usability at a $30–40 production cost. Programmed core AI features in C++ and C on an ESP32-S3 MCU, integrating DACs, microphones, LVGL, GPIO, and I2C for real-time interaction Targeted $100 retail price with 60%+ gross margin; planned crowdfunding and web launch