Freelancer profile translated to English.

Description

I build production data pipelines and AI agents for French SMEs and mid-sized companies.

Specializing in Data Engineering and Generative AI, I focus on two types of missions: structuring and industrializing data (ingestion, transformation, orchestration) and deploying intelligent systems, business RAG, autonomous agents, and LLM orchestration.

My relationship with tech is not academic. As a child of Terminator, Alien, and Matrix, I grew up with the idea that intelligent machines are not science fiction. They are built, line by line. This is what drove me to Data Engineering and AI: the desire to build systems that actually do something, not just slides.

What makes me concretely different: I built and deployed Kirari, my own AI agent in production, with a Qdrant vector memory, custom MCP tools, and a self-improvement loop. A truly living system that I use daily. It's this experience that I bring directly to your projects.

The data and AI ecosystem evolves fast, too fast to rest on its laurels. I continuously train myself, follow releases, and test new tools as soon as they emerge. If a technology from your stack is not on my list, it's not a blocker: learning quickly and well is part of my job, not my excuses.

My typical missions: data and AI audits (mapping, quantified roadmap), deploying RAG on your internal documents, building custom business agents to automate your complex workflows.

I work on a fixed-price basis, with a deliberately limited number of clients to ensure a real level of commitment on each project.

Full Stack:

• Data & Orchestration: Python, Prefect, Apache Airflow, PySpark, Kafka

• Storage & Databases: PostgreSQL, Qdrant, Elasticsearch, MongoDB, Cassandra, Neo4j

• AI & Agents: Claude API, MCP, RAG

• Infra & Deployment: Docker, Kubernetes, Scaleway

Industry field of expertise

Languages

French
Native or bilingual
English
Fluent

Workplace preferences

Remote only

Primarily works remotely

BOVIS
Data engineer
DIGITAL AND IT
February 2026 - Today (6 months)
Bondoufle, France
Design and complete implementation of a large-scale geospatial data architecture, covering 28 European countries. Responsible for the entire data cycle: modeling, ingestion, deduplication, semantic enrichment, and preparation of the AI layer.

Ingestion & pipeline

• Ingestion and normalization of ~50M POIs from 5 heterogeneous sources (Overture Maps, Foursquare OS Places, OpenStreetMap, Wikidata, GeoNames)

• Multi-format processing: GeoParquet (DuckDB + S3), PBF (pyosmium), JSON-LD, SPARQL, TSV

• Workflow orchestration with Prefect, optimized bulk inserts, idempotence and rerun management

Modeling & infrastructure

• PostgreSQL 16 + PostGIS + pgvector: partitioning by country, GIST spatial indexing, materialized views

• Split-DB architecture: self-hosted POI database + Supabase EU (Users/UGC) with sync layer and cross-database snapshot pattern

• Sovereign EU infrastructure, GDPR compliance, 12 SQL migrations applied, 315 unit tests
Deduplication & quality

• Multi-strategy deduplication pipeline: Wikidata QID identity pivot (conf. 0.99), Overture × Foursquare cross-referencing (0.95), fuzzy geo-semantic matching pg_trgm + ST_DWithin

• Review queue for sub-threshold decisions, complete audit log of decisions
Recommendation & AI

• Geospatial semantic recommendation engine: multi-criteria scoring, spatial filters, personalization by traveler profile

• Multilingual BGE-M3 embeddings pipeline for vector search and semantic matching (pgvector)

• Design of a conversational AI agent (LangGraph + Claude) with RAG architecture on the POI repository
PostgreSQL Python Prefect RAG ETL