Andersen is hiring an ML/MLOps Engineer in Germany for a project building a cloud-native AI platform and delivering scalable machine learning solutions for the healthcare industry.
The project is focused on building a cloud-native MLOps platform for the healthcare sector to support large-scale machine learning and AI workloads. It includes developing ML infrastructure based on Kubeflow, enabling LLM fine-tuning, traditional machine learning, and scalable data processing in a secure, zero-trust environment.
Responsibilities:
- Building and orchestrating ML pipelines using Kubeflow Pipelines (KFP v2).
- Training models on GPUs, including GPU resource management within Kubernetes.
- Fine-tuning transformers/LLMs, tracking experiments and models via MLflow.
- Building classic ML models (XGBoost, CatBoost).
- Working with data using SQL Server and DuckDB as a lightweight OLAP solution for efficient in-cluster processing of large datasets.
- Developing in Python (pipelines, integrations, tooling based on uv).
- Ensuring code quality: testing, CI/CD (GitLab CI).
- Working within a zero-trust / secure-by-default environment (network policies, restrictive container rights).
Kogo poszukujemy?
Must-haves:
- Experience as a MLOps Engineer / ML Engineer for 5+ years.
- Hands-on experience with Kubeflow Pipelines (KFP v2).
- Experience training models on GPUs.
- Experience fine-tuning LLMs/transformers.
- Experience with MLflow (model tracking).
- Experience with boosting models (XGBoost, CatBoost).
- Deep proficiency in the Python ecosystem and modern engineering practices.
- Experience working in regulated/enterprise cloud-native environments.
- Experience with SQL and large-scale data processing.
- CI/CD experience (GitLab CI preferred), clean code and testing practices.
- Level of English – from Intermediate+ or above.
- Level of German – from Upper-Intermediate or above.
Nice-to-haves:
- Pre-training experience for LLMs (beyond fine-tuning).
- Experience with GPU orchestration in Kubernetes.
- Experience in zero-trust environments (network policies, restrictive container rights).
- Experience with modern Python tooling (uv).
