name:       Florian Barre
role:       Data Scientist / AI Engineer  ·  Apprentice @ AFD (Agence Française de Développement)
location:   Paris, France
education:  ESILV — M2 Data & Artificial Intelligence (2021 → 2026)

mindset: >
  Innovation-driven engineer combining data engineering, DevOps and AI
  to design robust, scalable and performant technical solutions —
  with a strong analytical mind and a fast adaptability to complex challenges.

focus:
  - Production-grade AI applications & retrieval pipelines
  - RAG systems, vector search, multi-agent architectures
  - Full-stack engineering (FastAPI + React + PostgreSQL + Qdrant)
  - DevOps: containers, CI/CD, image scanning, hardened deployments
  - Quantitative research & financial market modeling
  - IoT, home automation, robotics, self-hosted infrastructure

currently_building:
  - Internal LLM applications & advanced RAG platforms
  - Quantitative backtesting engines (HMM regimes, factor strategies)
  - Smart-home ecosystem on ESPHome + Home Assistant
  - Custom Python utilities (LoggerPlusPlus, ConfigPlusPlus, TaskBrain)

Languages

AI · LLM · RAG

Data & ML

Back-end

Front-end

DevOps & Infra

IoT · Hardware · Robotics

Tools

Production-grade retrieval systems on large unstructured corpora — not just vectorstore.similarity_search().

• Hybrid retrieval — dense + sparse embeddings, BM25, vector search, RRF fusion
• Reranking — cross-encoder reranking on the top-k for precision boosts
• Indexing — Qdrant with hybrid indexing, payload filtering, multi-collection orchestration
• Document pipelines — parsing, OCR, semantic chunking, cleaning, normalization, structuring of unstructured data
• Metadata — automatic enrichment, classification, taxonomy mapping
• Architectures — multi-agent systems, tool calling, structured outputs
• APIs — OpenAI, Mistral, Hugging Face, vLLM, Ollama (local + remote inference)
• Optimization — inference cost/latency tradeoffs, caching, prompt engineering, evals

🧠   AI & RAG platforms — Production-grade chatbots and document assistants on top of Qdrant, FastAPI and modern LLMs, with hybrid retrieval and cross-encoder reranking.

📊   Quantitative research — Portfolio backtesting engine with HMM regime detection, factor strategies (momentum, low-vol, quality, value, carry, size) and HRP-based allocation.

📈   Financial pattern detection — Transformer-based time-series classifier on harmonic chart patterns, with temporal cross-validation across two decades of market data.

🏠   Smart-home ecosystem — Custom ESP32 / D1 Mini devices running ESPHome, integrated into Home Assistant with hand-tuned LED effects, sensors and Lovelace dashboards.

🖥️   Self-hosted infrastructure — Proxmox cluster running VMs and LXC containers for everything from game servers to dev environments.

🧰   Personal Python toolbox — LoggerPlusPlus, ConfigPlusPlus, TaskBrain, used across all my projects for consistent logging, configuration and scheduling.

🥇   Master Dev France — Finalist (2025)  ·  National programming contest focused on constrained optimization: logistics, path optimization, combinatorial problems. Local search, simulated annealing, fast iterative scoring.

🤖   French Robotics Cup — 3 editions  ·  Embedded development (Python / C++), low-level UART / SPI with custom CRC protocols, real-time schedulers, LiDAR-based autonomous navigation.

🛡️   CoHoMa project — Autonomous robot for the French Army  ·  Distributed ROS2 architecture, custom nodes / services / interfaces, DDS / QoS tuning, mapping & localization, multi-service Docker deployment.

👔   Team & project lead @ DaVinciBot  ·  Project management, technical recruitment, team supervision and evaluation.

📈   ENS Data Challenge  ·  Financial modeling and analysis on a real-world dataset.