Lead AI Engineer

Who we are.

We are a team of passionate experts with a clear ambition: applying digital technology to advance what matters for our clients and society.

Together we create reliable and responsive digital foundations for the world’s businesses, institutions, and communities.

Learn more on Advancing what matters.

Atos is hiring a Lead AI Engineer to build and operate the model, infrastructure, and data layer of the Belgium AI delivery practice. The role trains and fine-tunes custom large and small language models for project use, builds and administers the AI infrastructure those models run on, and leads the data management, modelling, and integration work that every AI project depends on.

Key Responsibilities

• Build and train custom LLMs and SLMs. Fine-tune, align, and adapt models for project needs.
• Create and administer the AI infrastructure. Set up compute, serving, orchestration, autoscaling, and cost controls.
• Lead data management for AI projects. Build pipelines, vector stores, feature stores, and lineage tracking.
• Lead modelling decisions across the portfolio. Choose the right approach: foundation model, fine-tune, or custom SLM.
• Lead AI integration into project codebases. Connect models with project and client systems through APIs, retrieval layers, and agent frameworks.
• Run the evaluation harness for trained models. Define metrics, test sets, and CI-based quality gates.
• Partner with the AI Governance and Platform Lead on governance. Provide model cards, data lineage, evaluation results, and prompt documentation.
• Mentor delivery-team engineers. Support teams on model selection, fine-tuning, and data-quality practices.

Required Qualifications

• Bachelor’s or Master’s degree in Computer Science, Machine Learning, Engineering, or equivalent experience.
• 6+ years in ML engineering, deep learning, or AI platform engineering.
• 3+ years training, fine-tuning, or deploying LLMs in production.
• Experience with model tuning techniques such as LoRA, QLoRA, instruction tuning, RLHF, or DPO.
• Knowledge of model-serving runtimes and GPU infrastructure on a major cloud platform.
• Strong data engineering background across pipelines, vector stores, feature stores, and lineage tracking.
• Senior-level Python skills with PyTorch and modern AI tooling.
• Good understanding of evaluation methodology, hallucination measurement, and retrieval failure analysis.
• Production experience taking at least one significant model from training to live operation.

Preferred Qualifications

• Experience training small language models or running continued pretraining.
• Hands-on RAG experience at scale, including chunking, embeddings, reranking, and evaluation.
• Experience with sovereign EU AI stacks for model training or serving.
• Hands-on with agent frameworks in production environments.
• AI FinOps experience covering cost control, right-sizing, caching, and workload optimisation.
• Open-source contributions, papers, or talks are a plus.

Technical Skills

• Modelling. PyTorch, Hugging Face, DeepSpeed, Accelerate, vLLM, TGI, LoRA/QLoRA, DPO, and RLHF.
• Inference and serving. vLLM, TGI, Triton, Ray Serve, SageMaker, AWS Bedrock, autoscaling, and latency/cost optimisation.
• Data. Databricks, Snowflake, vector stores, feature stores, knowledge graphs, and lineage tools.
• Cloud and compute. AWS, Azure, GCP, Kubernetes, Docker, GPU infrastructure, and sovereign EU stacks.
• Evaluation. Langfuse, Arize Phoenix, Ragas, DeepEval, Promptfoo, OpenAI Evals, and custom evaluation harnesses.
• Languages. Senior-level Python plus TypeScript, Go, or Rust for production integration.
• Engineering practice. Typed boundaries, testing, CI/CD, and infrastructure-as-code.

Learn more about us:

At Atos, we embrace diversity as the ultimate engine of ingenuity for our clients, and we constantly strive to create a culture where people feel supported and encouraged. Read more about our commitment here .

Whether it is fighting climate change, promoting digital inclusion, or ensuring trust in data management - tech for good sits at the core of our identity. With numerous global recognitions for our ESG practices, we are committed to building a better future for all by harnessing the power of technology. Learn more here.