Cloud & AI Architect

Building with AI,
writing about what works.

15+ years architecting cloud and AI systems — from Doctolib's infrastructure to award-winning AI products. Here I benchmark local LLMs, build agents, and document the techniques that actually hold up.

Read the writing → Projects

Currently

AI & DevOps Architect

Stealth startup

Building

AnvilTerm ↗

Open-source agentic terminal

Previously

Cloud Architect

Doctolib

Recognition

3rd / 9,500+

MS Azure AI Hackathon

01 — Latest writing

All posts →

Aug 24, 2026

Qwen3.8-27B on a 24GB Mac: 3 Kernel Panics and the 17.76 GiB Limit Apple Doesn't Print

Your Mac has 24GB of RAM. Metal reports 17.76 GiB as the amount the GPU can use without hurting performance. Comparing a GGUF file size against total RAM is the wrong arithmetic, and that gap is where the kernel panics live. Measured on one M4 Pro, including the runs that failed.

qwenllama.cppunsloth

Apr 5, 2026

Gemma 4 26B Won't Fit on My 24GB MacBook — Until I Did This

Ollama gives 2 tok/s with broken tool calling. I got 49 tok/s with perfect tool calling using Unsloth Q3_K_XL + llama.cpp. Then I built a Claude Code clone on top of it.

gemma 4llama.cppunsloth

Apr 3, 2026

I Tested Every Gemma 4 Model Locally on My MacBook — What Actually Works

Audio ASR in 3 languages, image understanding, full-stack app generation, coding, and agentic behavior -- all on a MacBook M4 Pro 24GB.

gemma 4local llmmacbook

02 — Recognition

3rd place among 9,500+ developers — Microsoft Azure AI Hackathon

Foodi Copilot, an AI nutrition assistant built on Azure AI and Cosmos DB, placed in the top 0.2% of one of Devpost's most popular hackathons and was featured on the official Microsoft Developer Blog.

Microsoft Blog ↗ Live demo ↗

FEATURED INTERVIEW — MICROSOFT HACKATHON WINNERS

03 — Selected projects

All projects →

Building with AI, writing about what works.