[floatingCatty]

Why

The old AcademicPages CV and cached homepage snippets were still easy to find. The live /cv/ was also showing an old experience list, missing current research, and using an unrelated visual style. Meanwhile an internal citation-audit Markdown document was accidentally live/indexable.

This is a canonical-profile cleanup, not a site redesign.

What changed

• Refresh /cv/ at the same permanent URL, replacing the obsolete archive/template page with a current, readable bilingual academic CV: education (McGill Physics PhD under Prof. Hong Guo; Tianjin M.E. 2024; Tiangong B.S. 2021), research roles (Nanoacademic 2025–26, AI for Science Institute 2022–24, BAAI 2020–21), research areas, DeePTB/VBCSR/AD-NEGF/flower, computing background and McGill fellowship. Publication entries are generated from _publications/, so future papers won't leave the CV stuck on an old list.
• Bring the /about/ page and custom /404.html into the existing lamp design; link the About page to primary research/software work, current citation records, and the canonical CV. User-facing design inherits the existing site's typography.
• On the existing homepage, add one low-key, readable line identifying the McGill Physics PhD and focus on electronic structure, quantum transport and scientific computing, with a direct research link. Preserve the lamp visual and original literary introduction. Add a link to the actual /about/.
• Keep the familiar old paths /resume, /resume/, /cv.html, and /about.html as noindex redirect stubs to the proper canonical page. This preserves human and crawler access without keeping duplicate CV pages.
• Remove docs/publication-source-audit.md from the public Jekyll build/sitemap only using an explicit exclude. The original maintainer document remains on GitHub with stable source links. The root Google verification file is untouched.
• Keep actual paper PDFs in files/, which are research resources, not outdated resume copies. Refine the AcademicPages navigation still used on standalone publication pages. Add a current branded 404 for truly deleted pages.
• Expand rendered-site QA: canonical, one H1, page description, ProfilePage JSON-LD, no retired CV text, working legacy redirect stubs, unpublished internal notes, preserved verification token, and sitemap hygiene.
• Document the actual path inventory and Search Console cleanup steps in .github/SEO_LEGACY_PATHS.md.

Important limits

The original cached Feb 2024 (expected) Google snippet can persist after deploying a corrected page; repository changes cannot delete a search engine cache immediately or erase past Git history. After deployment, use GSC URL Inspection on /cv/, /about/ and /. If an old snippet remains, GSC Removals → Clear snippet in search, not Temporarily remove URL (which would hide your valid current CV).

GitHub Pages redirects from jekyll-redirect-from are instant HTML meta-refresh redirects, not true HTTP 301. We keep one self-canonical content page per identity URL.

No ORCID or OpenAlex ID has been added without the author's confirmation. Work history is based on the recently approved CV; check exact public phrasing before merging.

QA

• Pages-compatible GitHub Actions Jekyll build plus scripts/audit_site.py.
• Review desktop/mobile CV and About; print the CV once; click language controls and old aliases.
• After merge confirm /docs/publication-source-audit/ responds 404 and /googlec07fc31230407367.html remains accessible; monitor GSC after Google recrawls.

Google references: redirect/canonical guidance, Removals / clear snippets, Jekyll redirect-from.

Verified build

GitHub Actions run 38186630514 passed on the latest commit (b3cc6ec): GitHub Pages Jekyll built successfully and 215 rendered-site checks passed across 10 research notes and 9 publications. Assertions cover the new canonical profiles, historical noindex redirects, absence of the accidentally published audit notes, 404 handling, Google verification token and machine-readable research metadata. Manual desktop/mobile visual QA and Google re-crawling are still outstanding.

Update: CV source review and full-language profiles

• Reconciled the CV against the owner's uploaded 2-page SLAI CV: Tianjin degree is M.E. in Electronic Information (not Computer Science); added research background and verified project scope across DeePTB/SLEM, VBCSR, RESCU++, RSATB (clearly work in progress), Nanoacademic/QTCAD, AI4S-Benchmark, honors and HPC/AI stack.
• Matched scholarly contact to the McGill email shown in the reference CV. Deliberately did not publish the phone number or upload the CV PDF itself.
• Added two missing publisher-verified npj Computational Materials papers: DeePTB-NEGF nanoelectronics, 11:375 (2025) and phonon-assisted optical spectra, 12:268 (2026), preserving the official journal author lists and linking only the underlying DeePTB background (not conflating these papers with other preprints).
• Published genuine /zh/cv/ and /zh/about/ alternatives. The existing /cv/ and /about/ remain English and self-canonical; pages expose reciprocal hreflang, semantic lang, locale-specific text, and visible non-JS language links. This is intentionally not a blind copy of the entire site.
• Automated build and content checks now cover 11 publications and two independent profile languages. Latest passing run: 38187406308 (268 checks passed).

Source note: the attached CV remains a private reference; no binary CV or personal phone number is added to this GitHub PR.