MLOps & Kubernetes Platform Engineer | LLM Agents, AI Infrastructure, HPC

1/2025 – offen

Architect and lead developer of ExaMLOps, the AI production platform for the EuroHPC-JU project SEANERGYS, deployed and running at a national supercomputing centre. UNIBO work-package task lead.

- Designed the end-to-end MLOps architecture: Prefect training pipelines, MLflow model lifecycle (versioning, metric-gated promotion, drift detection), Ray Serve multi-model inference, MinIO artefact storage.
- Built the full observability stack: Prometheus, Grafana, Loki and Tempo across training and serving, with alerting on model and infrastructure health.
- Delivered three control surfaces over the same API: REST, a first-class CLI, and a natural-language interface backed by an LLM agent.
- Ran the platform across hybrid HPC and Kubernetes targets, including SLURM workload integration and GitLab CI/CD for reproducible deployment.

Continuous Integration, DevOps, Machine Learning Engineer, MLOps, Python-Programmierer, Docker, Linux Entwicklung, Cloud Computing, Kubernetes

1/2024 – offen

Architect and lead engineer of KubeIntellect, an open-source multi-agent system that turns natural-language requests into safe, auditable Kubernetes operations across the full API surface.

- Designed a supervisor/worker agent architecture on LangGraph, with a stateful execution graph, PostgreSQL checkpointing and resumable runs.
- Enforced safety by construction: four-tier role-based access control, human-in-the-loop approval on every mutating operation, AST-based validation of generated code, and full audit logging.
- Built a Code-Generator agent that synthesises and registers new tools at runtime, so the system extends its own capability without redeployment.
- Implemented the backend in Python with FastAPI behind an OpenAI-compatible API, a LibreChat front end, and an SSH-tunnelled cluster access layer.
- Deployed to Azure AKS with Terraform, NGINX Ingress and Grafana dashboards.
- Published as peer-reviewed research and released under a dual open-source/commercial licence.

Postgresql, DevOps, Machine Learning Engineer, Python-Programmierer, Docker, Kubernetes

8/2022 – offen

Research and engineering on AI-driven operation of large-scale HPC and Kubernetes infrastructure, contributing to several EU Horizon projects (REGALE, DECICE, Graph-Massivizer).

- Built anomaly-detection and predictive models over production HPC telemetry, combining deep learning, graph machine learning and statistical methods on time-series data at cluster scale.
- Developed digital-twin and power/thermal models used for energy-aware scheduling and sustainable-computing decisions.
- Integrated the models into MLOps pipelines and an observability framework spanning HPC and Kubernetes, so detection runs continuously against live telemetry rather than offline traces.
- Created AOBench, an open benchmark for LLM agents performing HPC operations, with fidelity-validated environment snapshots (SLURM, Prometheus, energy APIs) and a scoring method that hard-fails any access-control violation.
- Co-supervise two PhD students and have co-supervised five MSc theses; peer reviewer for international journals and conferences.

Data Science, Machine Learning Engineer, Python-Programmierer, Linux Entwicklung, Kubernetes