open to remote work
→whoami
ai systems engineer
I build AI-native product stacks from scratch — multi-agent platforms with persistent memory, sovereign self-hosted inference, and real-time voice. Fifteen years of shipping behind it.
→ls work/ --sort=recent
An AI personal-assistant platform with persistent memory, real-time voice and multi-agent orchestration — shipping to web, iOS, Android and messaging channels. I architected it from scratch as founding lead engineer.
An OpenAI-compatible gateway, on-premise inference cluster and cross-client memory plane — removing external vendor dependency for every web, desktop, mobile and API consumer.
A custom ERP for a micro-insurance provider — policies, member records, payments and field operations in one system.
Replaced a phone-and-email facility booking process with a mobile-first application, wired into LINE for confirmations members actually read.
A flexible-benefits claim and reimbursement system handling balance calculation, document workflow and generated claim paperwork.
Property, agent and client management for a Thai real-estate agency, built on Laravel Nova for a small team that had to maintain it themselves.
An online/offline loyalty and marketing platform on Ethereum, co-designed with a team of three.
A values-first professional platform — a showcase built around what someone is working toward rather than a skills list. Founded from zero and shipped on web and mobile.
→cat practice.md
agent-systems/
Stateful agents that hold context over time — persistent memory blocks, multi-tier memory routing, context-window compaction, and approval-gated tool execution for anything that touches a user’s real data.
sovereign-inference/
Running your own models on your own hardware, with an OpenAI-compatible surface so nothing downstream has to care. No external vendor in the critical path.
real-time-multi-platform/
Voice agents over WebRTC, and one product surfaced across web, mobile, desktop and messaging — with the latency work and observability that keeps it usable.
engineering-leadership/
Available as a fractional CTO or advisor: architecture decisions, hiring, delivery process, and the build-versus-buy calls that are expensive to get wrong. Ran a six-person department for four years.
→git log --author="maxime" --oneline
Building an AI personal-assistant platform from scratch — persistent memory, real-time voice, multi-agent orchestration and gamification, shipping to web, iOS, Android and messaging channels.
Independent consulting on information systems and operations.
Ran the engineering department of a digital marketing agency for four and a half years — a six-person team and 30+ client sites in production.
A consultancy helping companies optimise processes and prototype ideas with the right technology per project.
Founded a startup from zero with a team of three.
Production support for a worldwide e-banking and cash management platform.
First engineering role, building the internal client portal.
→cat about.md
austcham thailand, bangkok
French, based in Bangkok, building software in Asia since 2010 — Shanghai, then Paris, then Thailand.
I have been the developer, the founder, the department head, and the person explaining to a board why the rebuild would take longer than sales promised. That range is the point: the hardest part of an AI product is rarely the model — it is the memory design, the failure modes, and the judgement about what should never be automated at all.
Right now that means agent architecture, self-hosted inference, and the unglamorous plumbing that decides whether any of it survives a real workload.
education
MSc Software Engineering
EPITA · Paris · 2005—2010
languages
away from the keyboard
→contact --all
Working on something where AI has to actually hold up? Tell me what you are building and where it is stuck.