'The design system that powers this website and will power my other projects. 29 Astro components across six families, a complete set of tokens (color, typography, spacing, radius), and a light/dark theme managed via CSS variables.',
context:[
'Designed for clean, technical interfaces: mono fonts for headings, sans-serif for body text.',
'Distributed under the npm scope @unkn0wndo3s, automatically published via a CI pipeline.',
'License: free to use, resale of the design system as a standalone product is prohibited.',
],
contributions:[
'Architecture of tokens and components (Custom Elements, typed events, slots).',
'Gitea Actions CI/CD pipeline for automated npm publishing and documentation deployment.',
],
stack:[
{icon:'astro',label:'Astro'},
{icon:'typescript',label:'TypeScript'},
{icon:'sass',label:'Sass'},
{icon:'gitea',label:'Gitea CI'},
],
links:[
{label:'Nova Design System v1.2.1',url:'https://nds.louis-potevin.dev/',icon:'astro'},
'A comprehensive self-hosted infrastructure: Git hosting, secret management, and application services, all containerized with Docker and continuously deployed.',
context:[
'Fully containerized with Docker, orchestrated, and reproducible.',
'Certificates and DNS managed through Cloudflare.',
'Services cleanly exposed using a reverse proxy.',
],
contributions:[
'Setup of a self-hosted Gitea instance with a Postgres backend and action runners.',
'A file organizer tool that currently runs in Java on Windows. It is being converted to Rust to achieve cross-platform Linux compatibility and optimize file search performance.',
context:[
'Current version: operational in Java, targeting Windows.',
'Gradual porting to Rust for performance gains and low-level control.',
],
contributions:[
'Rewriting sorting logic from Java to Rust.',
'Setting objectives for Linux compatibility and file search optimization.',
],
stack:[
{icon:'rust',label:'Rust'},
{icon:'linux',label:'Linux (upcoming)'},
],
links:[],
},
{
slug:'llm-tooling',
designation:'05',
title:'LLM Training & Inference',
tagline:'Small language model trained on a Reddit conversation dataset.',
'A language model training and inference project running on a dedicated machine. The model is trained on a custom dataset of Reddit conversations. Development is ongoing.',
context:[
'Execution on a dedicated workstation optimized for training and inference.',
'Dataset: conversation threads sourced from Reddit.',
],
contributions:[
'Assembling and preprocessing the conversation dataset.',
'Setting up the training and inference loop, currently iterating on results.',