Nassim B.

Syntree is a hybrid artificial intelligence platform combining LLMs, autonomous agents, a deterministic logic engine, and a Kubernetes microservices architecture.

Context

Creation of the Syntree platform, designed as a "code-first" AI where each request first goes through a deterministic logic engine before any call to an LLM.

Objective: design a scalable, secure, and industrialized infrastructure for several SaaS applications (CoreAI, Vigilis, Pepetes, Pharmalink, MyAccount).

Tasks performed

🔧 Architecture & Development

Design of a Rust microservices architecture (Axum) + Vue.js 3 + PostgreSQL + Redis.

Development of several backend services:

auth-service, notification-service, pepetes-api, vigilis-api, etc.

Implementation of a deterministic logic engine orchestrating tool/agent/LLM calls.

Creation of the LLM Service module integrating Qwen 3 Coder 30B + Whisper via Apple Silicon.

Setup of an internal AI agent framework (analysis, actions, validation, planning).

Development of Vue 3 frontends:

CoreAI, Vigilis, Pepetes, Pharmalink, MyAccount.

☁️ Infrastructure & DevOps

Setup of a Kubernetes cluster (k3d) with Traefik, NATS JetStream, Redis, PostgreSQL 16.

Complete GitOps architecture via ArgoCD (18 applications: DEV + PROD).

Self-hosted macOS GitHub Actions build pipelines (tests, lint, Docker, push).

Full automation of deployments:

git push → build → push Docker Hub → ArgoCD Image Updater → auto-deployment.

Kubernetes rolling updates with no downtime (2–5 minutes from push to production).

Secret management via Sealed Secrets.

🔐 Security

Design of a secure authentication system:

JWT HS256, Argon2, RBAC, internal SSO cookies.

📊 Observability & Quality

Complete drafting of the 2025 technical roadmap (OAuth2, Kong, Stripe Billing, SDKs…).

Setup of unit tests Rust/TS, E2E (Playwright, Q2 2025).

Implementation of the CI/CD validation plan (100% automated).

🎯 CONTEXT

The Collaborative Development (DevCo) project aims to streamline operational data analysis and reporting practices. Historically, several business stakeholders used heterogeneous tools (R, Python, Excel VBA). The objective is to centralize, secure, and modernize the processing of DOAAT's IS data. More recently, the project has expanded with the goal of providing businesses with generative AI tools: implementing RAG architectures, an MCP server, and dynamic chatbots integrated into applications.

🌐 DevCoStore Project

Management of about ten applications with integration of new solutions

Taking over existing applications developed in R or Python, and creating applications from scratch based on business needs using Python FastAPI and Vue.js.

Full project cycle: 🤝 business workshops → 📝 design (diagrams, databases) → 💻 frontend/backend → 🗄️ PostgreSQL → 🔧 GitLab CI/CD → 🚀 OpenShift deployment.

Migration from Python 3.9 → 3.13.

Development of REST APIs with Go (Fiber) (-> backend performance).

🤖 EDF IAG Project - Generative AI

Setup of an MCP server (FastMCP) to expose endpoints as tools

Development of a Vue.js conversational UI with dynamic display of contextual blocks

Design of an intelligent layout engine based on LLM response type.

Structuring responses via LLMs (GPT, Mistral) with metadata injection (JSON Schema).

usable by AI agents.

Implementation of an RAG architecture (Faiss) with augmented generation.

Integration of open-source LLMs (LLaMA 3, Mistral) self-hosted via Ollama.

Multi-agent orchestration with CrewAI (Researcher, Formatter, Answer).

📱 Mobile Project – eDOAAT

Design and development of a cross-platform mobile application (iOS/Android) with React Native.

💻:

Python, Rust, FastAPI, Go, Fiber, Vue.js 3, PostgreSQL, Kubernetes, GitLab CI/CD, LangChain, LangGraph, RAG, MCP, Ollama, Docker.

Within the digital department, participated in several cross-functional projects aimed at optimizing existing systems and creating new solutions.

️ Implementation of a cross-functional log management solution: Application architecture design and definition. Centralization and monitoring of cross-functional applications. Setup of log management (connection, processing, errors). Development of automated reporting and dashboards for quality monitoring.