Open to opportunities

vincenzo amoroso

@vincenzoamoroso

Message

I build secure, evaluated LLM agents and data products for constrained environments.

France

Message

What I'm looking for

I'm looking to build secure, production-grade AI and data products where reliability, privacy, and measurable outcomes matter—especially in constrained or sensitive environments.

At EDF, I build LLM agents and copilots for closed environments where sensitive data stays inside the information system. My agent selects tools, chains up to four calls, and answers from an offline RAG index spanning roughly 13,700 internal documents.

I also lead data and AI architecture work at EDF, including an address-matching workflow that raised industrial-consumption coverage from 30% to 90%. For BETer, I consolidated ten data sources, rebuilt a national vehicle fleet from 77 million registration records, and replaced Tableau Desktop dependencies with 13 standalone HTML reports.

Previously at SFR, I estimated and validated an 8% reduction in antenna-fleet electricity consumption through sector-level shutdown windows without observed quality-of-service degradation. I automated ARCEP measurement processing, built more than 40 Tableau dashboards, and used approved scraping to analyse billions of partner-network samples.

I build reliability into AI products through golden sets, non-regression benchmarks, CI checks, least-privilege access, and privacy-first architectures. My Nuclear Energy Copilot combines RAG, text-to-SQL, and layered query security, while my email-analysis application keeps mailbox data on the user’s workstation.

Experience

Work history, roles, and key accomplishments

Current

Nuclear Energy Copilot

Current

Personal Project

Jan 2026 - Present (8 months)

Developed an assistant in production returning sourced answers over a closed set of NEA/IAEA reports using hybrid vector and lexical retrieval on pgvector. Secured text-to-SQL with read-only validation, transaction timeout, and a no-write PostgreSQL role, and instrumented evaluation with a golden question set and CI checks.

RAG Pgvector Text To SQL PostGIS Python fastAPI Docker CI CD LLM Evaluation Security

Current

Data & AI Engineer, technical lead

Current

EDF

Oct 2024 - Present (1 year 11 months)

Designed and led the development of a conversational agent over internal and public data, using an internal LLM with tool selection and chaining, backed by an offline RAG index. Framed data and AI architectures for sensitive data and improved address matching coverage from 30% to 90%.

BI & Data Engineer

SFR

Sep 2021 - Sep 2024 (3 years)

Estimated and validated an 8% cut in antenna fleet electricity consumption through custom shutdown windows. Automated ARCEP measurement trace processing with outlier detection and developed over 40 Tableau dashboards.

Python Tableau Gitlab Data Processing Anomaly Detection Automation Dashboarding SQL Data Analysis

Full Stack Developer

GRT

Apr 2021 - Aug 2021 (4 months)

Developed a Python GUI with Redis integration for real-time data management, containerised and deployed via Docker. Wrote user documentation and contributed to installation of new network equipment.

Python Redis Docker GUI Development Documentation

Web Application for Email Analysis

Personal Project

Built a privacy-first web application for email conversation analysis, grouping messages into threads and rendering them as trees. The server is a stateless OAuth proxy with no database; analysis and rendering happen in the browser, with a multi-provider AI proxy and sandboxed iframe for untrusted HTML.