Marketplace Explore 6,000+ AI-powered automations to scale your business.

MLOps, GenAI & LLM Production

Retail – MLOps & GenAI Production

Industrialization of ML and LLM products on GCP with Kubernetes, CI/CD, security and observability.

Priority models put into production

More models assessed, ready for production

From POC to production

ML and LLM products industrialized on GCP, observable and secured

Project Details

Client

ADEO / Leroy Merlin

Sector

Retail – MLOps & GenAI Production

Year

2025

Key Technologies

PythonPyTorchGCPKubernetesDatadogVaultGitHub ActionsZenML

Mission

Senior MLOps Lead - GenAI in Production

Industrializing ML & GenAI from POC to Production on GCP

Hybrid assignment, France. English.

ML & GenAI productionization: take Data Scientists' POCs/MVPs all the way to production on GCP. Observability & reliability: logs, metrics, traces, SLI/SLO/SLA with Datadog; alerting & runbooks. Security & DevSecOps: secrets management (Vault), image hardening, compliance policies. Orchestration & CI/CD: Kubernetes (containerization, autoscaling), GitHub Actions workflows, artifacts & stacks (ZenML). Performance & FinOps: latency/availability optimization, streamlining of inference and infra costs. Documentation & operations: rigorous technical docs (HSV/TSD), operations playbooks, coding best practices. Model quality: data/model drift monitoring, performance KPIs, unit & integration tests.

Contexte & Environnement

Sector: Retail & e-commerce. Business context: putting multiple LLM-powered products into production and tracking their performance and metrics. Key use cases: (1) automated product description generation; (2) RAG-based chatbot; (3) scraping pipeline mapping financial signals/indicators via Google search APIs; (4) automatic SWOT matrix generation from financial data; + other POCs/MVPs. Stack & constraints: Kubernetes on Google Cloud; Vault for secrets; Datadog for logs/metrics/traces; strong security/observability and best-practice requirements (SLI/SLO/SLA); extensive documentation; access via VPN & SSO; English as working language. Stakeholders: Business, IT, Data, Ops; MLOps/Machine Learning Engineers team as a cross-functional interface. Project team: agile MLE/MLOps squad, rituals and daily collaboration with product & data teams.

Team

Senior MLOps Lead in an agile MLE/MLOps squad, with product & data teams

Environment

Hybrid, France – Kubernetes on Google Cloud, Vault, Datadog, VPN & SSO

Objectifs Cles

1Productionization at scale: 5 priority models put into production; 5 additional models assessed/remediated and ready for productionization.

2Standardized observability: global framework ready (acceptance) covering logs, metrics, data quality, SLI/SLO/SLA.

3Robust platform: end-to-end CI/CD pipelines, reproducible deployments/rollback, health monitoring on K8s.

4Performance & costs: optimized latency and availability; reduced inference and infrastructure costs (FinOps).

5Security: centralized secrets, compliance with internal policies, end-to-end auditability.

Technologies & Infrastructure

PythonPyTorchVertex AIOpenAIGCPKubernetesDockerGitHub ActionsZenMLDBTKafkaDatadogVault

Ready to Achieve Similar Results?

Let's discuss how we can drive measurable impact for your organization.

Other Business Cases

Explore how we drive results across industries

AI-Powered Video Research Platform

Media & Deeptech

Agentic R&D platform automating experimentation, evaluation and video-to-video generation pipelines.

Learn more

AI Skills Assessment at Scale

Banking

Skills ontology, large-scale assessments, ML scoring and personalized learning pathways.

Learn more

AI Architecture for Deliverable Automation

Rail Industry

Data and AI architecture to automate deliverable production, quality control and traceability.

Learn more