Applied AI Product Engineer

Building Sovereign, Secure & Production-Grade AI Systems

Bridging the gap between cutting-edge AI capabilities—Multimodal, Agentic RAG, Autonomous Workflows—and enterprise-grade safety, sovereignty & ethics.

Agentic Systems Advanced Hybrid RAG Sovereign & Safe LLMs Structured Outputs Multimodal & Robotics
Engineering Philosophy

How I Build with AI — The 4 Engineering Pillars

Delivering measurable product value through robust architecture, security, and user experience

01. AGENTIC SYSTEMS

Agentic Orchestration & Workflows

Designing autonomous agents capable of planning, tool/API execution, and multi-step tasks with deterministic reliability and structured fallbacks.

02. ADVANCED RAG

Enterprise & Advanced RAG

Hybrid search architectures (semantic, vector, BM25, reranking, Knowledge Graphs) guaranteeing precise, zero-hallucination domain knowledge retrieval.

03. SOVEREIGN & SAFE

Security, Ethics & Sovereignty

Deploying sovereign AI (Local/Open-Source LLMs, private cloud), zero PII data leaks, hard guardrails, and compliance with European regulations (EU AI Act).

04. MULTIMODAL & UX

Multimodal & Real-World Interaction

Integrating vision and audio models with robotics gateways to extend AI beyond text into reactive, physical, and real-time interaction.

Engineering Layers

Technical Stack & Architecture Matrix

Production-tested tools and frameworks organized by engineering layer

Engineering Domain Technologies & Frameworks
Orchestration & Agents
LangGraph LlamaIndex CrewAI AutoGen MCP Protocol
RAG & Vector DBs
Qdrant Pinecone pgvector BM25 / Hybrid Search Cohere / BGE Rerankers
Sovereignty & LLMs
Ollama vLLM Hugging Face Fine-Tuning (LoRA) Llama 3 / Mistral / Kimi
Evaluation & Safety
LangSmith Ragas (Evals) NeMo Guardrails Presidio (PII) detect-secrets
Engineering & Product
Python TypeScript FastAPI Docker Multi-Stage Cloud (GCP / AWS)
Production Case Studies

AI Product Case Studies & Demos

Enterprise-ready intelligent applications built for real-world impact and strict safety

Live Demo Available
🌍

Automatic i18n JSON Translator

Problem: Software localization corruption on interpolation variables.
Architecture: FastAPI microservice, Pydantic Structured Outputs, Regex protection & atomic disk checkpointing.
Impact: 100% variable safety, 0 key loss, deployed on Cloud Run.

Structured Outputs UX Produit FastAPI
Detailed Documentation
🚀

AI Product Engineering Blueprint (AIPE)

Problem: Initial technical debt and security leaks in AI projects.
Architecture: Poetry isolation, Ruff linter/formatter, Mypy strict (100%), non-root multi-stage Docker & pre-commit gatekeeping.
Impact: Developer onboarding < 5 min, 64/64 tests passed.

DevOps & QA FastAPI & Docker Ruff & Mypy
Live CLI & Interactive Demo

Automated AI Watcher CLI (Wrapper_CLI)

Problem: Heterogeneous AI technological surveillance & uncontrolled LLM API costs.
How it works: Multi-source ingestion (Text, Files, Web URLs), automated LLM analysis, FinOps pricing engine (40+ LLMs), Tenacity exponential backoff & SHA-256 caching.
Impact: Instant multi-format reports (Console, JSON, Markdown), zero-cost offline demo mode.

Python CLI FinOps & Cost Control Multi-Source Ingestion Typer & Rich
🚀 Coming Soon
🧠

Multi-Agent Workflow Orchestrator

Problem: Complex multi-step task execution requiring dynamic tool calling.
Architecture: CrewAI & AutoGen agent network with WebSockets and FastAPI event loop.
Impact: Distributed decision making with deterministic state rollbacks.

CrewAI AutoGen WebSockets FastAPI
🚀 Coming Soon
👁️

Vision AI & Analytics Microservice

Problem: Multimodal document processing and visual anomaly detection.
Architecture: PyTorch & OpenCV pipeline integrated with Gemini Vision API.
Impact: Real-time OCR and visual semantic extraction.

PyTorch OpenCV Gemini Vision