Quentin Dubois
AI Engineer — Project Manager

I build AI systems that act.

Autonomous agents · end-to-end systems · models designed and trained.

Discover the profile
Profile

AI, from research to product.

I'm an AI engineer. My job is to design complete AI systems — not notebooks, but products that run in production.

I work as much on model design and training as on orchestrating autonomous agents able to reason, use tools and collaborate.

I've been programming since middle school — my first projects were Minecraft servers in Java. That passion for code, never lost, led me self-taught all the way to applied AI.

One obsession: rigor. Evaluated, observable and reliable systems, where every technical choice is justified by the value delivered.

~20%

of AI projects actually reach production. My job: building the ones that do.

Outside work

Hyrox, and a former semi-professional World of Warcraft player — the same demand for performance, discipline and continuous progress.

Expertise

What I build.

  • Agents & agentic systems

    Autonomous agents and multi-agent workflows: planning, tool use, RAG, memory and self-correction.

  • End-to-end AI systems

    From data to production: ingestion, inference, APIs and scaling.

  • Model design & training

    Architectures, training from scratch, fine-tuning and rigorous evaluation.

  • LLMOps & production

    Observability, automated evaluations, cost control and real-world reliability.

Skills
Python· Java· PyTorch· Transformers· GitHub· Hugging Face· LangGraph· Pinecone· pgvector· SQL· PostgreSQL· FastAPI· Vue.js· Quasar· Docker· Kamal· GCP· Claude Code· Prompt engineering
Soft skills
Autonomy· Self-taught· Driven· Curiosity· Rigor· Perseverance· Product sense· Needs analysis
Languages
French — native· English — TOEIC 800
Projects

Selected work.

    Speech-to-Text · LLM

    Call transcription & summarization

    Automatic transcription and summarization of prospecting calls. 100% adoption across the company, with over 300 uses per month.

    Speech-to-Text· LLM· Summarization

    AI editor · Agents

    AI-assisted writing platform

    A Cursor-like editor for writing: generates copy and newsletters from client materials (PDFs parsed via OCR), the web or audio interviews, with agents that fact-check the content. A dedicated mode also generates call scripts from a client brief, via an LLM fine-tuned specifically for the task and continuously improved (collecting the edits users make to the generated content).

    Agents· LLM· Speech-to-Text· OCR· Fine-tuning

    Agents · RAG · NLP

    Autonomous monitoring system & newsletters

    Autonomous agents run daily monitoring on topics created on the fly and turn them into newsletters: trend charts, vector analysis of topics and vectorization of the collected content — including PDFs parsed via OCR — to write newsletters and articles.

    Autonomous agents· Embeddings· Monitoring· Data viz· OCR

    Agents · Automation

    Business automation agents

    Development of various agents integrated into existing applications to automate workflows and business tasks.

    Agents· Automation· Workflows

    Data · LLM

    Company data enrichment

    Enrichment from reliable sources (e.g. actual revenue) via a Monte-Carlo tree search guided by an LLM evaluator, with OCR parsing of source PDF documents. Several million rows processed.

    LLM· MCTS· Enrichment· OCR

    Company analysis · Sales enablement

    Account Intelligence

    Generates reports of several dozen pages analyzing a company against internal criteria (AI, SaaS…) to target and guide sales teams. Nearly 10 orders and around €50k in revenue within the first month of launch, from clients including a major data company.

    LLM· RAG· Reporting

    Entity matching

    Company data matching

    Matches a list of companies against the internal database using LLM arbiters, with matches enriched by web search. 80% precision on matches (false positives included).

    LLM· Matching· Web search

    RAG

    AI use-case directory

    A repository of how-tos and use cases for AI and agents in business — about 2,300 cases referenced, sourced or not — with their associated benefits, queryable via RAG retrievers; source PDFs parsed via OCR.

    RAG· LLM· Knowledge base· OCR

    Data engineering

    Prospect database structuring

    Normalization, structuring and use of prospect databases of over a million rows.

    Data· Normalization· Pipelines
Explorations

Built from scratch, out of curiosity.

    From scratch · AI

    Chess-playing AI & LLMs

    A chess engine built from scratch, then experiments with chess-playing LLMs — search, evaluation and reasoning.

    From scratch· Research· LLM

    Systems

    Blockchain from scratch

    A blockchain implemented entirely by hand to understand its mechanics: blocks, consensus, transactions.

    From scratch· Blockchain

    Solidity · Security

    Smart contracts & DeFi

    Solidity smart contract development and sandwich-attack (MEV) experiments to understand on-chain security.

    Solidity· DeFi· MEV

    Java · Game dev

    Minecraft servers (Java plugins)

    Design and development of Minecraft servers via custom Java plugins.

    Java· Plugins· Game dev
Background

Experience & education.

  • 2023 — Present

    Data & AI Project Manager

    ITFacto

    First on a work-study contract, then full-time. Since joining, I have designed and built the company's entire suite of AI applications self-taught, from scoping to production. Initially alone on the Data & AI scope, I supported its growth into a 4-person team — coordination, scoping and skills development.

  • 2023

    Internship — Java Developer

    Banque de France — Over-indebtedness department

    Java development for the over-indebtedness department, within an Agile team — my first immersion in a structured, iterative development environment.

  • 2021 — 2024

    Engineering degree (MSc) — Artificial Intelligence major

    CY Tech

    AI engineering program, final year completed on a work-study contract at ITFacto.

  • 2019 — 2021

    Preparatory classes for engineering schools (MPSI then PSI*)

    Lycées Clemenceau & Franklin Roosevelt — Reims

    Mathematics, physics and engineering science (scientific CPGE).

  • 2019

    French scientific Baccalauréat (science track)

    Lycée François 1er — Vitry-le-François

    Science track.

Recommendations

What people say about my work.

We wholeheartedly recommend Quentin, who now brings significant seniority to AI roles and roles bridging AI and business in research, exploration, and implementation environments.

Nicolas Beaumont

Managing Director · ITFacto

Contact

Let's build something intelligent.

An AI project, an agent system to design, or simply want to chat? Drop me a line.

contact@elpopeye.fr