Project Overview
This site is both my portfolio and a continuously evolving AI Coding product. It organizes my positioning, project cases, résumé, and product thinking into a maintainable bilingual content system, rendered through a custom Hugo theme. Visitors can inspect the
live product and
public source code, while the repository’s research notes, plans, changelog, and commit history make each iteration traceable from request to implementation and validation.
Pain Points & Opportunities
A one-page résumé communicates experience quickly but cannot show how a project was framed, what I personally owned, which actions mattered, or where the delivery boundary sits. Generic portfolio templates can launch quickly, yet they often force different projects into the same narrative and become difficult to evolve as experience changes. AI Coding can accelerate implementation, but without context, specifications, and validation it can also introduce bilingual drift, template syntax failures, UTF-8 BOM issues, malformed URLs, deployment differences, and inflated claims. The opportunity was therefore not simply to “have AI make a website,” but to build a repeatable collaboration process that turns product judgment into reliable public output.
Goals & Metrics
The site first needs to help target readers understand my direction quickly and follow a clear path from the homepage case to categorized projects and detailed evidence. It also needs independently maintainable bilingual content, project metadata, and visual components so that adding a project does not require rewriting templates. Finally, each AI Coding iteration must pass research, planning, human review, production builds, and critical-path checks. Delivery is currently assessed through live pages, bilingual completeness, responsive behavior, successful builds, and public iteration records—not through unverified traffic, recruiting conversion, or productivity claims.
User Research
The work draws on three evidence sources: my actual education and internship background, original project materials that can be shared publicly, and issues observed in the site’s information hierarchy, mobile experience, and deployment stability. Each iteration reviews whether the homepage surfaces the strongest evidence first, whether project cards distinguish personal responsibility from team outcomes, whether detail pages state result status, and whether both languages retain the same information structure. No formal interviews or large-scale usability study have been conducted, so claims about reader comprehension and effectiveness remain hypotheses to test.
Solution Design
The information architecture follows a “position quickly—inspect representative evidence—explore projects—review full experience” path. The homepage presents positioning and a primary case before capability evidence, additional projects, writing, and résumé access. The project index uses language-independent keys to group core AI, system development, product research, AI Coding demos, and data-modeling work. Each project page combines a case-at-a-glance panel with an eight-part narrative covering the problem, role, actions, results, and reflection. Bilingual Markdown owns the content, templates own stable structure, and CSS variables plus shared components keep the visual system consistent.
Execution Plan
Hugo and custom Go Templates generate the static site, with CSS, small JavaScript modules, and SVG assets providing themes, responsive navigation, project visuals, and interaction details. AI Coding follows a Research → Plan → Implement workflow: inspect the relevant code and content, document risks and technical paths, define affected files and a checkable Todo list, then implement after approval. Each iteration is validated with a production Hugo build, critical-page assertions, UTF-8 no-BOM checks, and guards for template strings and URL whitespace. Deployment supports both a Vercel build script and a GitHub Actions-to-GitHub-Pages workflow, while production failures are converted into durable repository rules.
Results Review
The project has delivered a public bilingual responsive website spanning a homepage, résumé, grouped project index, detailed case studies, and blog, with light/dark themes, mobile navigation, and explicit project states. The public repository contains the custom theme, content system, dozens of research and implementation plans, a changelog, release checks, and two deployment configurations, making the evolution of the product inspectable. This demonstrates an end-to-end AI Coding path from personal content strategy to running software; systematic external usability testing, behavioral analysis, and recruiting-conversion evidence remain future work.
Iteration Reflection
The central lesson is that AI Coding quality is not determined by how much code can be generated in one pass. It depends on whether the problem is framed clearly, the context is accurate, the plan can be reviewed, and the result survives real builds and page-level validation. Content integrity matters as much as engineering stability: team outcomes and planned metrics must not be presented as personal delivered results, while encoding, template, and deployment failures should become automated checks. The next valuable step is to use real navigation behavior to refine hierarchy, add stronger project artifacts, and test whether visitors can reach the core cases more efficiently.
