I put LLM systems into production inside stacks that already run.
Applied AI Engineer · 8 yrs · Paris · FR/EN. Mission-critical systems (RATP: 2023 World Cup, 2024 Olympics, zero production incidents). Legacy platforms rebuilt into multi-tenant AI SaaS. Evaluated systems, not demos.
Trusted by fast-moving companies
Three things, done properly
Not a service menu. The work that actually fills my weeks.
LLM systems in production
Retrieval, prompts, and orchestration that survive real traffic, with cost, latency and failure modes accounted for before launch, not after.
Integration with existing systems
The model is the easy part. The work is the auth, the data boundaries, the migration path, and the team that has to own it afterwards.
Evaluation and reliability
A written definition of "correct", a measured baseline, and a regression suite. Without it there is no way to tell an improvement from a coincidence.
Recent success stories
RATP event management system
Led the development of a critical management system for the 2023 Rugby World Cup and 2024 Olympics, using React, Node.js, and AWS. Implemented real-time monitoring with Datadog and containerized deployment with Docker & Kubernetes.
Staffdesk – AI Workforce Intelligence
Rebuilt an internal HR analytics platform into a scalable B2B SaaS, using LLMs, React, and PostgreSQL. Designed AI-powered dashboards and centralized data pipelines to eliminate inconsistencies and enable smarter decisions.
Keytt – Music Artist Growth Platform
Built a platform helping artists track data, improve visibility, and connect with creative experts. Led full-stack development with React, Node.js, and PostgreSQL. Integrated AI features for content planning and strategy using LLMs.
VibeCraft – AI-Powered Clipping Tool
Led development of a platform that auto-transforms Twitch/Kick streams into TikTok-ready clips using AI. Features highlight detection, creator dashboard, viral templates, and Stripe subscriptions.
Gateway: Creator credentials platform
Developed a full-stack Web3 platform with Next.js and Solidity, featuring NFT minting, wallet integration, and real-time notifications. Built with TypeScript, Material-UI, and blockchain tools like Ethers and WAGMI.
Most of the work happens before the code
By the time I write the first line, the hard part is usually already decided: what the system actually does, what the data allows, and what nobody wrote down.
Mapping the context
Reading the existing system, the data, and the constraints nobody wrote down. This is where the real problem is usually found.
Writing code
A smaller part of the project than people expect, and deliberately so. The code is the cheap part once the problem is clear.
Internal tooling
Evaluation harnesses, scripts, dashboards. What makes the result reproducible after I leave.
A team that goes straight to the code pays for it later, in the part of the project that is hardest to schedule.
Applied AI Engineer: production LLM systems
I'm Farouk, an applied AI engineer with 8 years building systems that run in production. I put LLM systems inside enterprise stacks that already work: evaluated, integrated, and handed over.
My background spans mission-critical systems (RATP: 2023 World Cup, 2024 Olympics, zero production incidents) and the rebuild of a legacy platform into a multi-tenant AI SaaS. Evaluated systems, not demos.
Where I have worked
RATP
Mission-critical event systems. 2023 Rugby World Cup, 2024 Olympics, zero production incidents.
Capgemini / Sogeti
Enterprise engineering.
Ministère de l'Intérieur
Security-cleared work.
Engineering school.
Trusted by great teams
Two people who worked with me on the same product, for a year.
Lionel Fabert
Founder & Creative Director – Keytt
Farouk is the invisible but fundamental architect of Keytt. Thanks to his technical vision and rapid execution, he transformed our creative ideas into a concrete, stable, and intuitive product. His AI integration not only added real value for our users but also helped differentiate Keytt in a saturated market. Working with him means moving three times faster.
Bakary Doucouré
Digital Strategy Director – Keytt
Farouk doesn't just code; he thinks about product, strategy, and scalability. From the start, he knew how to align business vision with technical constraints to make Keytt a powerful tool for artists. His ability to anticipate needs, create robust solutions, and always push for a better user experience was decisive in our launch.
Something stuck between demo and production? Tell me about it.
The stack, what already works, and where the last 20% stops. That is usually enough for a useful first answer.
Write directly
No form, no funnel, no booking link. One address, and the message lands in my inbox.
What happens next:
I read the context and ask what I actually need to know
An honest read on feasibility, including when the answer is no
Scope what it would take to get it into production
No pitch, no pressure