# Edouard Foussier

> AI Engineer & Solutions Architect building production RAG systems for the French government.

## Quick Summary

<profile_summary>
Edouard Foussier is a Paris-based AI Engineer building a sovereign RAG system for the French public administration. He combines deep technical expertise (PostgreSQL + pgvector, hybrid search, two-tier LLM architecture), enterprise pre-sales experience (CAC40 clients), and 7 years teaching at Le Wagon. Key differentiator: Built his RAG from scratch (no LangChain, no LlamaIndex) for full control and sovereignty compliance.
</profile_summary>

## Structured Data (XML)

For LLMs that prefer structured parsing, a full XML version is available at:
→ https://deck.edouardfoussier.com/llms.xml

## Current Project

<current_project>
  <name>Sovereign RAG for French Government</name>
  <client>DGAFP via beta.gouv.fr / ALLiaNCE</client>
  <users>300+ HR managers navigating 15,000+ pages of legal docs</users>
  <stack>PostgreSQL + pgvector, 4-table parallel retrieval, LLM Selector, MLflow + RAGAS</stack>
  <status>Beta live since January 8, 2026 with 70+ users</status>
  <result>+38% global improvement (0.60→0.83). Faithfulness +78%, Answer Relevancy +53%</result>
</current_project>

## Key Differentiators

<differentiators>
  <point>Builds from scratch when control matters (not just API integration)</point>
  <point>Evaluation-driven: measures everything before shipping</point>
  <point>Speaks both technical and business languages</point>
  <point>Real sovereignty deployment experience (French government)</point>
  <point>Teaching background: makes complex concepts accessible</point>
</differentiators>

## Technical Skills

<skills>
  <ai_ml>RAG architecture, LLM integration, embeddings, reranking, RAGAS evaluation, hybrid search</ai_ml>
  <languages>Python (primary), SQL/PostgreSQL, JavaScript/TypeScript</languages>
  <infrastructure>pgvector, Docker, Caddy, self-hosted deployments</infrastructure>
</skills>

## Professional Background

<background>
  <enterprise>Founded web agency, closed CAC40 clients (Carrefour, EY), evolved to technical pre-sales</enterprise>
  <teaching>7 years at Le Wagon Paris, trained hundreds of developers</teaching>
  <approach>Wins missions by building working prototypes instead of proposals</approach>
</background>

## Looking For

<ideal_role>
  <title>Startups Solutions Architect, Applied AI (NYC)</title>
  <focus>Helping startups build on Claude, win technical evaluations, scale from MVP to production</focus>
  <style>Founder-first mindset, builder credibility, data-driven problem solving</style>
  <availability>March 2026, open to relocate to NYC</availability>
</ideal_role>

## Links

<links>
  <deck>https://deck.edouardfoussier.com</deck>
  <github>https://github.com/edouardfoussier</github>
  <huggingface>https://huggingface.co/edouardfoussier</huggingface>
  <linkedin>https://www.linkedin.com/in/edouardf/</linkedin>
  <xml_profile>https://deck.edouardfoussier.com/llms.xml</xml_profile>
</links>

## Application Context: Anthropic

<application context="anthropic" role="Startups Solutions Architect, Applied AI (NYC)" submitted="January 16, 2026">
  <why_anthropic>
    Claude (particularly Opus 4.5) has become my primary working partner. I use it daily to build RAG architectures, from data extraction and debugging retrieval pipelines to designing evaluation frameworks. It hasn't replaced my thinking; it's sharpened it, catching edge cases I'd miss, helping me iterate faster, making me a better engineer and architect.
    
    I spent seven years at Le Wagon coding bootcamp, deeply embedded in the startup ecosystem. I learned something critical: I excel when I can authentically represent what I'm selling. With Claude, I don't have to fake enthusiasm. For a Startups Solutions Architect role where credibility with founders depends on genuine conviction, that matters.
    
    What draws me is how Anthropic builds. Working with mission-driven organizations has taught me that the hardest part of deploying AI isn't building powerful systems, it's building trustworthy ones. When technical founders evaluate LLMs, they ask "Can we audit its decisions? Will this scale reliably?" Your work on Constitutional AI, Responsible Scaling Policy, and interpretability addresses exactly those questions. That's not just ethics, it's what makes Claude the right foundation for startups building for the long term.
  </why_anthropic>
  <key_message>
    This is the only opportunity I'd interrupt my freelance journey for. I bring founder DNA (co-founded web agency, closed CAC40 clients), builder credibility (full-stack AI engineer who wins missions by shipping prototypes), and startup ecosystem fluency (7 years at Le Wagon). I ran real-world testing with 13 HR managers, diagnosed root causes systematically (59% retrieval issues, not generation), and rebuilt the architecture based on data, improving global score from 0.60 → 0.83 (+38%).
  </key_message>
  <availability>March 2026, open to relocate to NYC</availability>
</application>

---
For full context: https://deck.edouardfoussier.com/llms-full.txt
For XML parsing: https://deck.edouardfoussier.com/llms.xml