Technical Articles
Long-form articles published on Medium. Most start from a practical problem and come with a working repository you can run yourself. Summaries here; full articles on Medium.
33 articles shown
Why iptables-based networking struggles as clusters grow, and how Cilium uses eBPF to provide identity-aware L3–L7 policies, better performance and much deeper visibility into pod-to-pod traffic.
Machine learning needs its own operational layer. The article walks through orchestrating parallel training, model selection and auto-scaled serving on Kubernetes with Kubeflow Pipelines and KServe.
Addresses the “two-world” problem of apps in Kubernetes and infrastructure in separate Terraform pipelines. Covers Crossplane's XRDs, Compositions and Claims, how they fit GitOps, and a practical Azure example.
A LangGraph supervisor that lets an LLM choose between heuristic scoring, database lookups, threat-intelligence and web-search tools to classify streamed log events and raise alerts.
Extends retrieval-augmented generation to video: indexing spoken content and visual scenes, then retrieving and streaming only the segments that answer a natural-language query.
Fundamentals of semantic search and vector databases, followed by a working support-ticket search service that stores embeddings in Qdrant and exposes them through FastAPI.
The hands-on half: provisioning each secret provider with Terraform and injecting secrets into Kubernetes through the External Secrets Operator.
Why secrets should not live in code or plain manifests, and how external secret managers change the way applications on Kubernetes receive credentials.
When CPU and memory are the wrong signals: scaling consumers on queue depth in SQS and Kafka, and on CPU load, using KEDA scaled objects.
Releasing safely with canary releases and A/B testing driven by Istio traffic management and Flagger, including promotion decisions based on custom metrics.
Making microservices fault tolerant with Istio: advanced circuit breaking applied to a real-world scenario, then chaos engineering with fault injection to prove the system recovers.
Completing the pipeline: a Schema Registry for consistent, evolvable records, and a sink Kafka Connect cluster that writes the change stream from Kafka to MinIO as Parquet for analytics.
Hands-on: building a source Kafka Connect cluster and connector image with Strimzi on Kubernetes, capturing PostgreSQL changes into Kafka, and testing the CDC pipeline end to end.
Why change data capture matters for real-time data, typical use cases, and how Strimzi and the Kafka Connect architecture make CDC manageable on Kubernetes.
Taking the same Dagster pipelines to AWS: infrastructure provisioned with Terraform and the orchestrator and code locations running on Amazon ECS.
An MLOps workflow in ClearML: projects, MinIO for artifact storage, versioned datasets, and turning dataset preparation and training tasks into a reusable pipeline.
Running multiple Dagster code locations on Kubernetes with Helm: each pipeline as its own deployment, the webserver and daemon wiring, and how scheduled runs become Kubernetes jobs.
A step-by-step deployment of two ETL pipelines to Dagster with Docker: reading from S3, transforming data and loading it into PostgreSQL through Dagster IO managers.
Orchestrating Dask's distributed computations from Dagster, with a machine-learning pipeline example that generates, splits and trains on data in parallel.
Writing faster, more readable Python: comprehensions vs generators, named tuples, arrays, memory views, deque and keeping sequences ordered with bisect.
Centralising authentication in front of Kubernetes services: the external-auth architecture at the Ingress, implemented with OAuth2 Proxy using Keycloak and GitHub as identity providers.
Enforcing cluster policy with OPA Gatekeeper: the API request lifecycle, ConstraintTemplates and Constraints, and examples such as required resource limits, trusted image registries and mutations.
Backup and recovery for stateful workloads: installing the VolumeSnapshot CRDs, backing up and restoring PVCs, and scheduling automated backups.
What makes a database highly available, why read replicas matter for disaster recovery and multi-zone setups, and how to configure streaming replication and promote a replica in PostgreSQL.
In Turkish: what Infrastructure as Code is, why Terraform, how it compares with Ansible, installation, core commands and a first resource example.
In Turkish: semaphores, wait/notify communication between threads and the producer–consumer pattern.
In Turkish: coarse- vs fine-grained locking strategies, synchronized blocks and ReentrantLock compared.
In Turkish: race conditions, critical sections, which operations are atomic in Java, and locking basics.
In Turkish: threads vs processes, scheduling, why multithreading, daemon threads, thread pools and executors.
In Turkish: why Elasticsearch, installation and configuration, core concepts and field mappings for log data.
In Turkish: configuring Logstash and parsing application logs with Grok patterns before sending them to Elasticsearch.
In Turkish: centralising application logs with the ELK stack, starting with log collection through Filebeat and the Beats family.
In Turkish: building custom validation annotations in Spring Boot, combining them, validation groups and payloads for error handling.