Kaiwei's Page

Experience

Where I've worked and what I've done.

  1. Software Engineer Intern · Canadian Museums

    May 2026Present

    Toronto, Canada

    • Implemented a Stripe-based checkout flow with temporary ticket reservation and automatic expiration release, preventing overselling while recovering unpaid reserved inventory.
    • Built QR-code ticket generation and staff scanner validation flows, enabling secure on-site admission checks with scan history tracking for auditability.
    • Developed refund and payout workflows with Stripe Connect, payout reserves, and ledger records, supporting financial reconciliation across orders, refunds, and organizer transfers.
    TypeScriptNode.jsStripePostgreSQL
  2. Software Engineer Intern · Ministry of Education

    Sep 2024Apr 2025

    Toronto, Canada

    • Reduced semantic search query latency by optimizing retrieval and indexing pipelines, including chunking strategies, relevance filtering, and vector search configuration.
    • Improved answer accuracy by 12% on a human-evaluated benchmark by refining retrieval quality, document preprocessing, and context selection for RAG responses.
    • Reduced deployment time by 80% by implementing Azure DevOps CI/CD pipelines with automated JUnit testing, achieving 90%+ test coverage across release workflows.
    PythonJavaAzureRAGCI/CD
  3. Software Engineer, Open Source · LangChain-AI

    Sep 2024Dec 2024

    Remote

    • Engineered performance optimizations for FAISS metadata filtering using lazy condition evaluation, achieving a 51% improvement in filter computation efficiency over the initial implementation.
    • Implemented context caching for the @langchain/google-genai package by integrating Google's File and Cache Manager APIs, enabling reuse of cached inputs across Gemini model invocations to reduce redundant token processing.
    PythonTypeScriptFAISSGemini API
  4. Software Engineer Intern · Changchun Construction Investment

    May 2023Aug 2023

    JiLin, China

    • Developed an executive dashboard with Streamlit, Pandas, and Spark tracking sales conversion and retention across 20k+ users, replacing manual Excel reporting and adopted by business units for weekly reviews.
    PythonStreamlitPandasSpark