Xebia

xebia.com

Xebia organizes its business around Applied & GenAI, Intelligent Automation, Cloud & Data Modernization, and Digital Products & Platforms. Buyers can engage it for AI roadmaps, custom AI development, production ML engineering, data platforms, software development, and selected cloud-security work. These are distinct service scopes: a strategy engagement does not necessarily include implementation, and platform or training purchases should not be treated as engineering engagements. Its customer cases provide concrete evidence of ML pipelines and cloud infrastructure delivery. ^[1] ^[2] ^[3] ^[4]

Services and delivery

AI strategy and governance advisory

Advisory. Assesses AI maturity, identifies and validates use cases, and develops roadmaps with timelines and milestones. Deliverables include governance policies, role definitions, and ethical guardrails; implementation requires a separately defined scope. ^[2]

AI and generative AI development

Implementation. Builds AI solutions from prototypes through production systems, including LLM fine-tuning. Supported applications include forecasting, churn prediction, fraud detection, enterprise knowledge assistants, document processing, structured extraction, and computer vision. ^[5]

Agentic AI development and integration

Implementation. Designs multi-agent systems and builds custom LLM-powered agents with tool use. Delivery includes enterprise interfaces, platform deployment, guardrails, monitoring, and adoption support. ^[6]

MLOps engineering

Implementation. Moves machine-learning models from experimentation into production through ML architecture, automated CI/CD, monitoring, and drift handling. The service also addresses maturity assessment and cross-functional lifecycle ownership. ^[7]

AI platform engineering

Implementation. Designs and implements tailored platforms for generative AI and conventional machine learning using open-source and cloud technologies. Scope includes LLM fine-tuning, LLMOps, and MLOps rather than platform licensing alone. ^[8]

AI managed operations

Managed services. Operates and optimizes AI systems through LLM pipeline monitoring, platform support, model-performance and data-quality management, infrastructure updates, incident response, and drift monitoring. ^[9]

Data engineering

Implementation. Designs and builds data architectures, pipelines, streaming systems, and lineage capabilities. The service explicitly includes Spark, Databricks, Flink, dbt, and OpenLineage; this offering alone does not establish AI model development. ^[10]

Software and platform engineering

Implementation. Develops digital products and platforms, modernizes legacy software, and provides quality engineering. The documented scope includes functional programming with Scala and Kotlin. ^[11]

IT strategy and enterprise architecture

Advisory. Provides technology advisory, architecture, governance, and application-estate assessment. Application portfolio rationalization helps buyers define changes to their technology estate before committing to execution. ^[12]

Robotic process automation implementation

Implementation. Discovers automation opportunities and develops, customizes, and implements RPA bots. Ordinary process automation is not classified here as machine-learning delivery. ^[13]

Cloud-security assessment and architecture

Advisory. Assesses cloud-security risks and designs secure, resilient architectures. This is a specific cloud-security advisory scope, not evidence of every cybersecurity specialty. ^[14]

Cloud-security implementation

Implementation. Implements cloud-security and resilience engineering measures, distinct from assessment and architecture advice. ^[14]

Agentic OS platform

Software product. A packaged enterprise AI orchestration platform built on OutSystems, with knowledge integration, governance, cost controls, and prebuilt components. The evidence does not establish standalone licensing terms. ^[15]

AI training and upskilling

Non-technical service. Provides role-based learning and certification programs covering generative AI, machine learning, LLMOps, and prompt engineering. Training is separate from consulting and implementation. ^[16]

Company details

Track record

For AXA Direct Assurance, Xebia reports jointly implementing a Databricks retraining pipeline for home-insurance price optimization, with MLflow tracking and registry, data and model versioning, Python quality improvements, and CI/CD tests. The cited growth and model-update improvements were expected outcomes, not established realized results. ^[3]

For Gumtree, Xebia reports implementing Google Cloud data infrastructure and a tailored MLOps platform, including Vertex AI for search recommendations and Pub/Sub, Dataflow, and Bigtable components. The case dates migration completion to the end of 2022. ^[4]

An unnamed apparel and lifestyle customer case describes an Azure-native Super Agent accessible through Microsoft Teams, with identity enforcement, role-based permissions, logging, and tracing. Further production-readiness hardening remained in the future-work scope. ^[17]

Engagement and delivery

The agentic AI offering separates ideation and readiness work from a six-week proof of concept and subsequent integration or platform deployment. This is an advertised engagement sequence, not a guaranteed schedule for every project. ^[6]

Scope and limitations

The supplied evidence establishes AI operations services but not contractual support hours, response times, or service-level commitments.
Agentic OS licensing, implementation boundaries, and commercial terms remain unverified.
Case studies demonstrate particular engagements, not guaranteed outcomes or production maturity for every solution.
Some follow-up research requests timed out; no claims here rely on the unread Kramp case or separate AI Platform Design & Implementation page.

Questions to ask

Which team owns discovery, implementation, production acceptance, and ongoing operations?
Which references match our model type, platform, industry, and regulatory requirements?
What acceptance tests cover accuracy, reliability, security, and model drift?
Which licenses, cloud charges, accelerators, and reusable components are priced separately?
What support hours, incident obligations, handover materials, and exit rights will be contractual?

Overview