Freelancer profile translated to English.
Back to original languageAbout Mathias
Generative AI Architect.
I design and industrialize agent and RAG systems for B2B companies operating under strict constraints: sensitive data, regulatory requirements (AI Act, GDPR, NIS2), need for auditability. For these players, an approximate AI is a risk, not a feature. I don't deliver a demo that impresses in a meeting: I deliver a system that works in production, traceable, controlled, and without imposed dependence on a non-European provider.
What I deliver
Agents in production with multi-role orchestration (LangGraph, Claude Agent SDK, OpenAI Agents SDK) and industrialized RAG: systematic source citation, anti-hallucination guardrails, evaluations, and monitoring. Migration to MCP. Full LLMOps: evals, guardrails, observability, cost control.
And the layer most people overlook: AI Act and GDPR compliance designed from the architecture, sovereign hosting (European VPS, your cloud, or SecNumCloud base), self-hostable models when independence from third-party APIs is required.
All documented and transferred, so your teams remain autonomous after I leave.
Stack
Python, FastAPI, LangChain, LangGraph, Pydantic AI. Claude, OpenAI, Mistral, including self-hosted. pgvector, Qdrant, Chroma. Docker on European VPS, your cloud, or SecNumCloud base. Next.js and Supabase Postgres when the system needs to deliver its own back-office.
I also act as technical director: architectural decisions, stack choices, code review, orchestration of a senior freelance team. You leave with a system that runs, clear documentation, and trained teams, not with an additional dependency.
Paris.
French, English.
Founder of Ceres Broker.
Computer Engineering graduate, Cybersecurity major (CY Tech).
French
Native or bilingual
English
Native or bilingual
Can work on-site
Paris (up to 50km)
Experience
- LeadcheeAI/RAG ArchitectTECHMay 2025 - September 2025 (4 months)Paris, FranceDesign and deployment of a next-generation CRM powered by RAG agents, where value lies not in data entry but in automated action. Agents analyze each incoming piece of content (email, form, external signal), enrich the prospect profile using internal and external sources, generate a prioritized action plan based on the commercial context, and trigger follow-ups at the right time.Core challenge: reliability in a noisy environment. Recommendations traceable to their source, safeguards against erroneous actions, controlled end-to-end latency. Prospect data processing designed with GDPR compliance in mind. Agents orchestrated in Python on a dedicated RAG base, exposed via a Next.js commercial interface.
- StudelyAI ArchitectTECHApril 2025 - February 2026 (10 months)Courbevoie, FranceDesign and deployment of two structuring AI systems. Tripled operational productivity in the concerned functions, without additional hiring.System 1Autonomous editorial pipeline. Orchestrated agents covering the entire chain: brief ingestion, factual source verification (anti-hallucination safeguard), SEO gap analysis, editorial structuring, multilingual translation, and automatic publication on Webflow. Removed the bottleneck that was blocking organic acquisition.System 2Augmented community management. Real-time intent classification agent that qualifies each message and directs it to the correct queue, with systematic human escalation for sensitive cases rather than blind automation. Analytics back-office directly feeding product decisions. User data processed in accordance with GDPR.
- Ceres BrokerFounder of an Engineering FirmSOFTWARE PUBLISHINGNovember 2024 - Today (1 year and 7 months)Paris, FranceEngineering firm specializing in the design and deployment of AI systems in production for French B2B companies, with compliance and sovereignty as basic requirements, not options.Three areas: internal operations automation (workflows, document processing, reporting), commercial acquisition and qualification systems, integration of AI agents into existing business processes. AI Act and GDPR compliance designed from the architecture; data hosted in Europe and remaining the client's property at all times.Fixed monthly subscription model, delivering results, not man-days, with a team dispatched as needed (AI, engineering, web, DevOps) and a dedicated AI hub for each client. Technical direction across the entire portfolio: architecture, stack choices, code review, recruitment, and orchestration of a senior freelance bench.
Recommendations
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- Computer Engineering Degree, Computer Science major, 2024 graduationCY Tech2024Diplôme d'ingénieur, Génie Informatique - Majeure Cybersécurité