Hi, I’m
Patryk.

For a digital AI avatar at a state-owned real estate company, I built and operated the observability stack on an on-premises, multi-tenant Kubernetes cluster running NVIDIA NIM, vLLM and Milvus. I deployed Prometheus, Alertmanager, Grafana, Loki, Promtail / Grafana Alloy and NVIDIA DCGM Exporter via Helm, with configuration in Git. Metrics and logs covered time to first token, inference latency, token throughput, errors, GPU utilization and memory, and Milvus and RAG pipelines. I configured PromQL alerting, Grafana dashboards and SLO tracking; provided namespace-aware visibility and integrated monitoring with the team’s RBAC, secrets and Network Policies. I applied NVIDIA NeMo Guardrails, evaluated model, prompt and RAG variants with MLflow, optimized Milvus retrieval, and analyzed capacity to recommend GPU allocation, replicas and scaling. I also co-developed, containerized and integrated LangChain / LangGraph agentic AI and RAG components into production workflows.

For a global premium automotive manufacturer, I developed, automated and validated a zero-downtime certificate-rotation strategy for HPE Ezmeral Runtime Environment, a Kubernetes platform for production analytics and AI workloads. I tested the approach in a non-production environment, prepared a runbook and rollback procedure, and implemented it in production. I also configured and stabilized identity integration, storage and health checks after installation.

Across projects, I built and maintained GitHub Actions and Jenkins pipelines for automated tests, container image builds and Kubernetes deployments. I packaged services and platform components as Helm Charts with environment-specific values. I implemented Argo CD deployments with declarative desired state in Git, automatic cluster synchronization, and traceable changes and rollbacks through Git history.

For an internal HPE product project, I co-developed a Python forecasting microservice spanning batch and streaming ingestion, training, deployment and inference. I ran training, experiment tracking and model versioning in the AWS-based HPE Machine Learning Development Environment (Determined AI), integrating Amazon S3 data sources. I exposed forecasts through FastAPI for HPE Private Cloud AI, LLM applications and enterprise systems, deployed the containerized service on Kubernetes, and contributed pytest unit and integration tests and Git-based code reviews. I also contributed to time-series and stream-processing architecture design with Apache Flink.

For a global automotive manufacturer, I developed a Python transformation engine integrating Apache Kafka with PostgreSQL and Neo4j analytics pipelines, with Docker containers for production. I co-designed and developed similarity search using client-specific Jaccard, Pearson and Spearman evaluation metrics to improve metadata retrieval. I applied Neo4j Graph Data Science for similarity, centrality and pathfinding, developed supervised metadata-classification workflows, administered PostgreSQL and Neo4j databases, performed exploratory data analysis and validated ML results.

For a professional sports organization, I developed and scheduled Apache Airflow DAGs for the data pipelines of a lottery-based fan-engagement platform. Retries and monitoring supported reliable, auditable processing.

For an international trading and distribution company, I developed scheduled Python automation for file retention, log rotation and server-side data lifecycle management. I also supported server upgrades, certificate renewals and runtime migrations.

For a national construction company, I conducted workshops using HPE’s Edge-to-Cloud Adoption Framework (E2CAF), mapping data capabilities, access models and RBAC models. I supported the establishment of a Transformation Business Office for governance and decision-making.

I participated in the HPE & NVIDIA AI Development Initiative covering AI infrastructure operations, GenAI, LLMs and multimodal AI, with technical training in HPE Private Cloud AI and Azure ML MLOps deployment patterns. For an LLM hackathon, I designed and developed an autonomous LangChain agent. I also gave technical presentations on AI-based aerial image analysis and HPE’s GenAI portfolio and participated in HPE DeliverAI Sessions.