半開 · Selected work & projects
10 years in tech, moving from software engineering into AI application development.
My work spans software testing, Python development, and product management.
I focus on AI agents, workflow automation, and how software is built in the age of AI.
Seeking roles in AI application or full-stack development, with an interest in AI testing and evaluation. Resume available by email.
半開
Let flowers bloom halfway
leave room for the rest
Projects
A look at the problems, implementation, and validation behind my work.
- 01Private Knowledge AssistantA locally run assistant for private knowledge: hybrid retrieval, streaming answers, and traceable sources.OverviewCase study
- 02WorldQuant Alpha Research ToolkitAn engineered workflow for WorldQuant BRAIN Alpha research: candidate generation, multi-metric evaluation, and human-approved submissions.OverviewCase study
- 03AI Application Development Study RepoLearning notes and milestone projects in AI application development: LangChain, RAG, LangGraph, evaluation, and engineering practice.Open sourceGitHub (opens in a new tab)
- 04HengcangA multi-currency household asset ledger: an Android app and a web workbench share one ledger service, tracking accounts, holdings, and cash flows across a three-bucket plan. The web workbench requires sign-in.LiveCase study
- 05Investment DashboardsTwo public read-only dashboards: index allocation and recurring-investment watch, plus A-share short-term money flow. Scheduled jobs refresh the data snapshots that ship with the code.LiveOpen dashboards (opens in a new tab)
- 06Character ProfilesA workspace for character creation material: text profiles, reference images, and shared pose or expression templates in one place, with optional local LLM help for prompt refinement and image generation.LiveOpen site (opens in a new tab)
Writing
Notes on what I build and learn.
- Building hybrid RAG retrieval with keywords, vectors and traceable citationsStart with a keyword baseline, then add real vector retrieval, rank fusion, access and revision checks, evidence budgets and citation validation using a synthetic example.
- From software testing to AI application development: choosing and validating projectsTurn API testing, test data and failure analysis into a complete AI project, with repeatable checks for success, failure and recovery and clear evidence boundaries.