I design and operate production-grade infrastructure for AI/ML workloads — from MLOps pipelines and Kubernetes platforms to computer vision systems running on the edge. 4+ years turning complex cloud architecture into measurable operational impact.
MLOps pipelines for production AI/ML models with SLA enforcement and drift management. CI/CD migration to Azure DevOps + Tekton, Medallion data architecture with CDC, IaC with Terraform/Pulumi, cloud networking, and full observability with Grafana Alloy + Prometheus.
LLM agentic system on LangChain + OpenAI models with ReAct paradigm, API-first CRM integration and pre-server MCP orchestration. Production observability via LangFuse.
BI & analytics microservices migration from VMware Tanzu to AKS with custom CI/CD; MySQL 5.4 → 8.4 database migrations.
Kubernetes-ready GTE real-time tracking system at Frankfurt Airport — eliminated manual apron runs, accelerated baggage offload. Model iteration YOLOv5 → RF-DETR, on-prem edge inference on Nvidia Jetson Orin at 3× RTSP streams, 25+ FPS.
Technical bridge between Italian client stakeholders and German development team — backlog, sprint planning, delivery sign-off. All in English (IELTS 8.5).
EfficientDet fine-tuned with TensorFlow to automate contract date extraction and signature recognition from scanned banking documents.
Azure · GCP · Oracle OCI
VPC / peering · Load Balancers
Cost optimization
Kubernetes · Helm · Docker
ArgoCD · Tekton · Kustomize
Terraform / OpenTofu · Pulumi
Model deployment · Drift mgmt
LangChain · LangFuse · MCP
Nvidia Triton · TensorFlow
YOLOv8 · RF-DETR · OpenMMLab
Edge inference · Jetson Orin
Real-time RTSP pipelines
dbt · Debezium · Datastream
Medallion architecture · CDC
PostgreSQL · BigQuery · Cube.dev
Prometheus · Grafana Alloy
Loki · SLA monitoring
Keycloak · IAM least-privilege
Open to conversations about platform engineering, MLOps, and everything cloud-native.