Projects

2026 Link to heading

Data Pipelines & Automation Link to heading

Job Market Intelligence Engine [private repo]

13 platforms scraped daily for tech job listings
~7,800 jobs aggregated across 2 DuckDB databases
3-stage CI/CD pipeline (lint → test → deploy) — 222 tests
MCP server exposing live data to AI agents
Python · dlt · DuckDB · FastMCP · pydantic · typer · loguru

Deep Research Agent [private repo]

Agentic LangGraph pipeline — 9-10 nodes : search → verify → synthesize
3 LLMs orchestrated (Gemini 2.5 Pro, DeepSeek V4, DeepSeek Chat)
Parallel fetching : 5 concurrent URLs with rate limiting (Semaphore)
Structured markdown reports generated automatically 3x/week (Docker cron on VPS)
170 tests · LangGraph · LangChain · Serper · httpx · Docker

Knowledge Base Automation [private repo]

Obsidian vault : 43,000+ documents — personal second brain
37,000+ wiki entries auto-generated by LLM
Hybrid search BM25 + vector (QMD) · GPU embeddings (ROCm, RX 7800 XT)
Multi-provider LLM (Gemini, Inception Labs)
Python · Gemini API · QMD · Obsidian

Web Scraping & Market Data Link to heading

Immo-Stras — Real Estate Aggregator GitHub

In production since 2025 : immo.vpdata.fr
LeBonCoin scraping with geographic filtering (Strasbourg, 5 zip codes)
FastAPI web dashboard + smart deduplication (exact + fuzzy matching)
Docker + Traefik + HTTPS deployment · CI/CD via GitHub Actions · n8n integration
FastAPI · SQLAlchemy · Docker · Traefik · n8n

Tesla Market Analyzer [private repo]

Automated market analysis : price, mileage, battery health tracking
237 unit tests with full database isolation (DuckDB/SQLite)
CI/CD via GitHub Actions + Discord notifications
17 CLI commands (Typer + Rich) : stats, best-deals, report
Python · DuckDB · Pydantic · Typer · GitHub Actions

Tools & Infrastructure Link to heading

Cognitive Framework [private repo]

Hybrid S1/S2 architecture for coding agents — 1,148 passing tests
6 project types auto-managed (web, api, data, cli, lib, bot)
MCP server with 7 tools (FastMCP/stdio) · installable profiles
LangGraph · LangChain · Pydantic · pytest · FastMCP

Second Brain Workflow [private repo]

Multi-source pipeline (YouTube, Medium, web articles) → structured Obsidian notes
5 development phases · ~10,000 lines of Python · 336 functions
AI summaries via Gemini 2.5 Flash (1M token context window)
Telegram interface : /batch, /enrich, /index commands
Anti-detection : Webshare proxy rotation, cookie management
Multi-language support with automatic fallback (fr → en)
VPS deployment with systemd · pipeline state tracking
Python · Gemini API · yt-dlp · Telegram Bot API · Obsidian

Find optimal insights for music band development
Tracking website visitors with matomo
Performance content evaluation on social networks (Youtube & Meta)
Give feedback and advices for strategic development\

Social Media Analytics Pipeline [private repo]

Unified ELT pipeline for YouTube + Meta : dlt ingestion, dbt transformations
OBT model ~2,300 lines in BigQuery, 4 datasets
GitHub Actions orchestration (3 daily runs)
Python · dlt · dbt · BigQuery · YouTube API · Meta Graph API

Tracking OTM festival Link to heading

Matomo Analytics installation on a personal VPS
Visit tracking
Sales analysis
…

Survey of performing arts professionals Link to heading

Collapse of the performing arts sector following COVID (2025)
Impact of COVID-19 on the live entertainment sector : - 80% of license renewals
Cleansing and analysis of 80,000 rows using BigQuery, SQL and Google Sheets
Data segmentation to isolate major industry players
Graphic visualizations: maps, graphs, tables (Datawrapper, Google Sheet)
Identification of dominant sectors and key user types (e.g. individuals vs. other natural persons)