AI Architect · Founder of IgnitionAI

I design and build production-grade Generative AI systems, RAG platforms, AI agents and enterprise AI infrastructure.

My work sits between AI architecture and hands-on engineering: retrieval systems, agentic workflows, MCP servers, evaluation pipelines, LLM integrations, observability and the infrastructure required to move GenAI applications from prototype to production.

I've worked on AI systems for BNP Paribas CIB, Sanofi, Brevo and SUEZ, while building open-source tools and products through IgnitionAI.

A reusable GenAI infrastructure platform for building document intelligence products and enterprise AI applications.

• Hybrid search, reranking, HyDE and configurable retrieval strategies
• RAG evaluation and A/B testing
• Tool-using AI agents
• Visual agentic workflow builder
• REST API and TypeScript SDK
• Native MCP server
• Embeddable AI widgets
• Multi-provider / BYOK architecture
• Enterprise data ingestion and knowledge management

The goal is simple: provide reusable AI infrastructure so teams don't rebuild ingestion, retrieval, evaluation, agents and integrations for every new GenAI product.

An experimental reinforcement learning framework for the JavaScript ecosystem, inspired by Unity ML-Agents.

The objective is to train agents directly from JavaScript / Three.js environments and export trained policies for deployment through ONNX.

My AI architecture and engineering company.

I work with companies on:

• GenAI & agentic system architecture
• Enterprise RAG
• AI agents and tool orchestration
• MCP servers
• Retrieval and RAG evaluation
• LLMOps / observability
• Enterprise AI platforms
• AI governance and production deployment

Architecture and industrialization work around enterprise GenAI adoption, including:

• Internal GenAI platforms built around OpenWebUI and Azure OpenAI
• Microsoft Copilot Studio agents
• MCP servers exposing enterprise tools and APIs
• RAG knowledge-base management and replication
• Retrieval evaluation with Precision@k, Recall@k and F1@k
• LLM-as-a-judge and A/B evaluation
• Agent evaluation campaigns
• Production guardrails and human-in-the-loop workflows
• RBAC, secrets/PII protection and environment governance

Built and deployed secure on-premise GenAI and RAG systems for large banking document corpora.

Worked on:

• FastAPI / TypeScript AI services
• LangChain & LangGraph
• pgvector and OpenSearch
• Advanced retrieval and metadata filtering
• Time-decay and freshness strategies
• Tool-augmented agents
• Source-grounded generation
• RBAC and auditability
• Langfuse / Phoenix observability
• On-premise LLMs

Designed and evaluated GenAI systems for pharmaceutical marketing workflows:

• RAG pipelines
• Retrieval optimization
• Ranking experiments
• ReAct agents
• Agentic RAG
• Automated generation and validation workflows
• Evaluation and observability

Worked on the industrialization of a GenAI platform for marketing automation, including LangGraph-based multi-agent orchestration and tool-using agents.

AI / Agents

LangGraph · LangChain · MCP · RAG · Agentic RAG · OpenWebUI · Copilot Studio · Azure AI Foundry · Hugging Face

Languages

TypeScript · Python · Rust · Java

Backend

Bun · Hono · FastAPI · Node.js · NestJS

Frontend

Next.js · React

Data & Retrieval

PostgreSQL · pgvector · Qdrant · OpenSearch · Azure AI Search · Cosmos DB

AI Engineering

PyTorch · TensorFlow · ONNX · Transformers

Infrastructure

Docker · Terraform · GitHub Actions · Azure DevOps · Azure · AWS · GCP

LLMOps / Evaluation

Langfuse · Phoenix · Weights & Biases Weave

I'm currently working with Packt Publishing on a technical book about building agentic AI applications with Next.js, TypeScript and MCP.

I also created GenAI Labs, a French training program focused on building Generative AI applications for TypeScript and Node.js developers.

I publish technical content about RAG, agents, MCP, Azure AI and GenAI engineering on YouTube.

Before specializing in AI engineering, I worked as a full-stack engineer across enterprise projects.

Before software, I served in the French Army's RICM, eventually reaching the rank of Corporal.

That path probably explains a lot about how I approach engineering today: build things that work, understand the system end-to-end, and take ownership of production.

I'm particularly interested in:

AI Agents · Agent Infrastructure · RAG · MCP · Context Engineering · Retrieval · LLM Evaluation · LLMOps · Enterprise AI · Reinforcement Learning