Freelancer profile translated to English.

Description

Stop letting your models die in the POC phase.

Hello 👋🏼 I'm Rahul, MLOps & Systems Engineer.

Production AI doesn't forgive: unacceptable latency, exploding Cloud costs, instability under load. I help Data teams and institutions transform their research work (Python/Notebooks) into robust, sovereign, and cost-effective inference infrastructures.

💡 What I bring to your teams concretely

Large-scale LLM Inference: Deployment and optimization of models via vLLM to handle massive loads (e.g., industrializing AI support for a major public administration).
Cost-Killing & GPU Optimization: Reducing latency and memory footprint through quantification, ONNX, and inference engine tuning.
Critical Infrastructures & Sovereignty: Building highly available Kubernetes clusters (EKS/GKE) and secure deployments on sovereign clouds (Outscale, AWS) using Terraform and strict CI/CD pipelines.
Deep Tech & Low-Level: Where the Python ecosystem shows its limits, I rewrite critical components in C++ or Go to ensure maximum system performance.

🛠️ Stack & Ecosystem

AI & Data: Python, PyTorch, vLLM, ONNX, FastAPI
Cloud & Infra: AWS, Outscale (SecNumCloud), GCP, Kubernetes, Docker
DevOps: Terraform, GitHub Actions, GitLab CI
Core & Performance: C++, Rust, Go

🎯 Why collaborate with me?

Most MLOps profiles are tool assemblers. I am a systems engineer.

In production, facing a memory leak or an incomprehensible bottleneck on your clusters, I'm not blocked by framework abstraction. I know how to go down to the low-level code to diagnose and fix what others can't see.

📩 Have a POC to industrialize, a sovereign infrastructure to build, or GPU costs to halve? Let's discuss!

Industry field of expertise

Languages

French
Native or bilingual
English
Fluent

Workplace preferences

Remote only

Primarily works remotely

Taker
Computer Vision & AI Engineer
CONSULTING AND AUDITS
December 2025 - Today (8 months)
Paris, France
Design and optimization of advanced inference pipelines for complex image processing problems.

Inference & State-of-the-Art (SOTA):Integration, optimization, and production deployment of ultra-precise segmentation models from the state-of-the-art (Meta's SAM2, RMBG).
Computer Vision Pipeline:Full development of the background removal pipeline with strong precision and computational performance constraints.
Microservices Architecture:Integration of AI pipelines into robust IT microservices infrastructures.
Microservices Architecture Computer Vision Machine Learning FastAPI Pytorch
Datakeen
Machine Learning Ops (MLOps)
CONSULTING AND AUDITS
April 2025 - August 2025 (4 months)
Paris, France
Managed the technical infrastructure and lifecycle of LLM & Anti-fraud models for over 20 major clients and public administrations.

Large-Scale & High-Performance LLM Deployment (vLLM):Orchestrated the production deployment and optimization of complex LLM solutions (e.g., industrializing a classification model for a major national administration). Ensured low latency and high availability by deploying these models via vLLM and FastAPI.
Critical Architectures & Sovereign Cloud:Met strict security requirements (data isolation, sensitive document processing) by designing and deploying custom architectures on multi-cloud and sovereign environments (AWS, SecNumCloud via Outscale/3DS, Scaleway), including managing GPU clusters under load-balancing (e.g., for a major insurance player).
Core Engineering & Anti-Fraud Optimization:Led the training, optimization, and deployment of Deep Learning models (classification, anomaly detection, OCR) for identity document anti-fraud, ensuring scalability for clients.
CI/CD Industrialization & Engineering Culture:Improved reliability and accelerated delivery pace for the engineering team by building automated CI/CD pipelines. Also wrote reference technical documentation (Standard Operating Procedures) to standardize deployment practices.
Real-time Observability & Reliability:Ensured the robustness of production systems by implementing advanced performance probes for continuous infrastructure monitoring, tracking model drift (data drift), and guaranteeing optimal operational margin.
This experience allowed me to deliver robust, scalable, and maintainable solutions in close collaboration with technical teams.
AWS Docker Kubernetes Pytorch CI/CD
Bal des fous
Backend & Cloud Engineer
ENTERTAINMENT AND LEISURE
January 2026 - Today (7 months)
Paris, France
Worked as a technical consultant on back-end architecture and engagement features for a social networking application.
Real-time Architecture (FCM):Designed and implemented a real-time push notification system via Firebase Cloud Messaging, built to efficiently handle user load and engagement.
Cross-Posting Engine:Developed and integrated a complex cross-platform publishing system ensuring data flow synchronization and reliability.
Google Cloud Microservices Architecture NestJs Back-End Development Software Architecture