I’m Quentin Erdinger. I live in France, where I build software from the cloud down to the metal.

I like owning a problem end to end — from cloud platforms and AI agent tooling down to interpreters and runtimes — and shipping something people actually use.

Most recently I spent six months at Adobe in Paris building a brand-intelligence engine. It started life as a web and PDF scraper; I turned it into an LLM-powered service that reads a brand’s websites and documents and extracts its full identity — products, pricing, tone of voice, layout, assets, campaigns and promotions — for Adobe’s Solution Consultants. It processed more than 200,000 items in its first month across 7+ regions. Replacing raw HTML with Playwright accessibility-tree snapshots cut LLM token usage by ~43% and latency by ~62%, and I integrated MCP so AI assistants like Claude could safely use internal tools and data.

Before that I wrote COBOL on z/OS for health cover systems at Euro-Information, rebuilt a legacy Java application in multi-tenant ASP.NET Core at Liebherr, and laid down Liebherr’s Azure foundations as a DevOps intern.

I studied at Epitech, earning a Bachelor and an MSc in Computer Science, and completed a dual-degree MSc in Advanced Computer Science at the University of Kent with Distinction. My thesis, supervised by Dr Stefan Marr, evaluated pointer tagging and NaN boxing for a Rust-based implementation of the SOM interpreter.

Outside of work I co-created Luxoria, a plugin-based photo editor whose open-source NuGet libraries have passed 10,000 downloads, and I took second place in AWS’s University Engagement Program with SimpleFlow, an AI assistant for GitHub workflows.

// If a function produces a weird outcome, it might not be a bug — just an undocumented feature.