Skip to content

Portfolio — LabML : ML in your browser (paire EN/FR + captures) - #22

Merged
dapiced merged 1 commit into
mainfrom
claude/portfolio-labml
Aug 21, 2026
Merged

Portfolio — LabML : ML in your browser (paire EN/FR + captures)#22
dapiced merged 1 commit into
mainfrom
claude/portfolio-labml

Conversation

@dapiced

@dapiced dapiced commented Aug 21, 2026

Copy link
Copy Markdown
Owner

Quoi

Nouvelle entrée portfolio couvrant app.dominicdapice.com de bout en bout, au format des paires existantes (front matter portfolio, lang, translation_url croisés) :

  • _portfolio/labml-ml-in-your-browser.md (EN) et _portfolio/labml-le-ml-dans-votre-navigateur.md (FR)
  • 4 captures dans assets/img/labml/ (panneau d'incertitude + analyse par segments, EN et FR — les images de tête)

Contenu de l'article

  • Le pitch : une plateforme ML complète — pas une démo devant une API — où chaque calcul tourne sur la machine de l'utilisateur (« ouvrez l'onglet Réseau pendant qu'elle entraîne huit modèles : rien ne sort »)
  • Le ML Lab (/ml) : profilage, détection de fuites, zoo de 8 modèles dont GBDT et MLP maison, et surtout l'évaluation honnête — baseline, intervalles bootstrap à 95 %, analyse par segments (figure titanic), seuils chiffrés par coûts, Shapley exact, comparaison de runs, export/réimport de modèles, exploration sans cible, prévisions Holt-Winters
  • Le Data Studio (/data) : score de qualité, recette rejouable, isolation forest, jointures aux orphelines nommées, contrôle de dérive PSI
  • Le terrain de jeu IA (/ai) : vision on-device (ONNX Runtime Web) et assistant de données déterministe, honnêtement étiqueté « pas un modèle de langue »
  • Sous le capot : algorithmes from scratch et seedés, Web Workers, 237 tests unitaires + 49 e2e, Lighthouse ≈ 0,99, PWA hors-ligne, CSP stricte — la confidentialité comme architecture, pas comme promesse

Generated by Claude Code

New portfolio entry covering app.dominicdapice.com end to end: the ML Lab
(/ml — honest evaluation, intervals, segments, thresholds, model reuse),
the Data Studio (/data — quality score, replayable recipes, joins, drift)
and the AI playground (/ai — on-device vision, deterministic data
assistant), plus the engineering story (hand-written seeded algorithms in
Web Workers, no backend, offline PWA). Two screenshots per language in
assets/img/labml/.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UKw6oNC8iZ9Kn7q6x4qom4
@dapiced
dapiced marked this pull request as ready for review August 21, 2026 06:26
Copilot AI lite review requested due to automatic review settings August 21, 2026 06:26
@dapiced
dapiced merged commit 7f41a5a into main Aug 21, 2026

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Adds a new bilingual (EN/FR) portfolio entry for “LabML — ML in your browser”, aligned with the existing _portfolio translation-pair pattern (lang + cross-linked translation_url) and including an image for social/preview metadata.

Changes:

  • Added new English portfolio page: _portfolio/labml-ml-in-your-browser.md.
  • Added new French portfolio page: _portfolio/labml-le-ml-dans-votre-navigateur.md.
  • Included cross-linked translation metadata and per-page image front matter for both entries.

Reviewed changes

Copilot reviewed 2 out of 6 changed files in this pull request and generated no comments.

File Description
_portfolio/labml-ml-in-your-browser.md New EN portfolio entry for LabML with translation link + preview image front matter.
_portfolio/labml-le-ml-dans-votre-navigateur.md New FR portfolio entry for LabML with translation link + preview image front matter.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants