AI products. Real operations.
End-to-end ownership.
I turn business problems into working software: guest-support agents, knowledge retrieval, evaluation tools, and the web and mobile platforms behind them. Sole developer of GnG Vacation's property and AI systems.
Agents, retrieval and review workflows built for daily operations. Each case study explains the problem, my implementation and the engineering decisions behind it.
Production AI · Human feedback
A tool-using agent handling approximately 10,000 guest messages a month. Property knowledge, web search and Lark escalation connect automated answers with staff expertise. I also built response scoring and feedback instrumentation.
RAG · Business workflows
An employee knowledge assistant with hybrid retrieval, plus AI-assisted CRM updates from Zoom call summaries. Vector, keyword and title search work together; confidence checks and audit records support lead classification.
Data provenance · Decision support
Compares short-, mid- and long-term rental strategies using Airbnb and Zillow comparables, listing history and financial tools. Reports distinguish sourced inputs, assumptions and missing data.
Multimodal AI · Human review
Compares cleaner uploads with reference photos to flag missing supplies and staging issues. Structured findings support pass, retake and manager-review decisions.
AI depends on the systems around it. I built the property data, channel integrations, payments and operational workflows that connect these products.
100+ properties · Sole developer
The shared foundation for calendars, pricing, property records, leasing inquiries and a unified inbox. Integrates Airbnb, Vrbo and Booking.com through Channex, with owner reporting and native mobile clients.
Payments · Channel integration
Availability checks, quotes and reservations connected to the PMS. Stripe authorization and capture flows include idempotent requests and payment-state handling; longer stays feed into CRM follow-up.
Calendar sync · Operations
Matches reservations to cleaners and sends Lark reminders for the coming seven days. The workflow checks calendars every two hours and detects changes so staff can respond to shifting bookings.
Customer experience · Discovery
A customer-facing rental website with destination pages, property discovery and inquiry flows connected to the wider booking and operations system.
SwiftUI and Kotlin clients bring the same operational data to property managers, owners and residents.
Native property management on iPhone and iPad: rental calendars, bookings, guest communication and operational dashboards connected to the same PMS backend.
Owner access to rental calendars, income reports, leasing progress and service requests. I also developed an Android owner app in Kotlin and Jetpack Compose.
Mobile access to leads, customer records and follow-up stages, extending the CRM workflow beyond the desktop.
A SwiftUI resident app connected to lease information, rent status, maintenance requests and community services through shared PMS APIs.
My route into software began with accounting and revenue analysis. Repeated questions about property income became analysis tools; scattered operational knowledge became a PMS, then AI and mobile products. Economics and public policy trained me to question the inputs and explain the outcome. Engineering lets me turn that analysis into a system people can use.
Master of Public Policy
gpa 3.986 / 4.0
BA Economics, Minor in Data Science
gpa 3.757 / 4.0 · Provost Honors 2020–2022
Software Developer · Previously Accounting & Revenue Analysis
Attorney Assistant
Assistant Project Manager
Beyond the build
Research, experiments and lessons on AI systems, human judgment and engineering. Available in English and Chinese.
Explore the blog ↗Briefings on AI, semiconductors and technology markets, with source links and a dated archive. Written in Chinese.
Read the briefings ↗Interested in AI applications, evaluation tooling and the engineering that turns models into useful products.
