Freelancer profile translated to English.

Description

AI Memory Architect.

I help companies that have deployed Claude/ChatGPT internally and realize their teams re-prompt from scratch every day, without collective knowledge building, without personal continuity.

My approach: implement a structured memory (pure Markdown, readable, auditable, portable) that the AI consults at each interaction.

No black box vector DB.

Result: real AI ROI, not just sleeping Enterprise ChatGPT licenses.

Stack: Claude Code + MCP, OpenAI Custom GPT, Next.js / TypeScript / Supabase, Python for pipelines. Open source IP in progress: PolyMind

Target missions: AI audit + LLM memory setup + hands-on training. Short one-shot model: 5-10 days per mission, clear deliverable, skills transfer. Remote preferred, bilingual FR/EN.

Ideal profile: CIO/CTO of companies with 50-500 employees that have deployed ChatGPT Enterprise or Claude internally and are experiencing disappointing ROI. Also: HR/transformation teams looking to train their employees in practical AI (not ChatGPT 101).

Languages

French
Native or bilingual
English
Fluent

Workplace preferences

Can work on-site

Paris (up to 20km)

Memvalt
Founder & Builder — Memvalt (LLM Memory SaaS)
March 2026 - Today (3 months)
Paris, France
SaaS for structured memory for teams deploying ChatGPT, Claude, or AI agents internally.
Memvalt transforms individual LLM usage into a collective leverage: business knowledge is
capitalized in a team vault consulted by each AI used by employees.

Multi-tenant architecture (Clerk Organizations), Stripe seats, Supabase database, deployed on Vercel. Stack
Next.js 16 / TypeScript.

Target: SMEs and scale-ups with 10-200 employees who have deployed ChatGPT Enterprise or Claude and are looking to avoid individual usage stagnation.

Live:
Next.js Typescript Supabase LLM AI Memory
Projet personnel — Open Source
Creator & Architect — PolyMind (Open Source)
February 2026 - Today (4 months)
Paris, France
Structured memory architecture for LLMs (Claude, ChatGPT, Cursor). Pure Markdown system,
readable and auditable — alternative to opaque vector DBs for companies adopting generative AI
internally.

Components: note taxonomy, Python indexing (Graphify), MCP (Model Context Protocol) integration for Claude Code, deterministic prompt structuring.

Objective: become the reference memory layer for companies deploying Claude Code, Cursor,
ChatGPT Enterprise. Open source method, consulting services offered in parallel
Claude Code LLM MCP Prompt Engineering AI Agent
Mission freelance — secteur retail
Fullstack Developer & API Integrator — ESL SaaS
January 2026 - Today (5 months)
Paris, France
SaaS in production for a retail client: third-party API integrations (Hipoink — BM ↔
BH synchronization), data synchronization pipelines, internal dashboard for operational piloting.

Stack: Next.js, Supabase, TypeScript. Effective production deployment, ongoing maintenance.
Next.js Typescript Supabase API Integration PostgreSQL