Thoughts and Writings

An introduction to evaluating large language model outputs: metric types, key dimensions (correctness, relevancy, hallucination, safety), when to choose model-based vs statistical metrics, and how to start evaluating your LLM system.

A practical guide to making large language model applications reliable and safe in production. Covers key metrics, tracing, quality checks, security signals, and cost control strategies.

An introduction to testing large language model systems: test types (unit, regression, safety, performance), practical workflow, common pitfalls, and how to get started building a test suite you can trust.

A practical guide to embedding security into DevOps: build automation, shift risk detection left, monitor in production, and grow maturity without slowing delivery.

Key kubernetes metrics to monitor for reliable, efficient clusters — from resource usage to control plane health and actionable alerts.

Explains canary deployments and how to implement progressive delivery in Kubernetes using tools like Flagger for safer, automated rollouts.

Explains canary deployments and how to implement progressive delivery in Kubernetes using tools like Flagger for safer, automated rollouts.

Looks at when and how to use namespaces to separate workloads safely within a kubernetes cluster. Explains their limits, common pitfalls, and why staging and production should be isolated beyond namespaces.

Practical guidance for building an internal platform on top of kubernetes. Covers core building blocks, ecosystem tools, observability and security guardrails, and trade-offs to expect when running at scale.

Explains the “platform as a product” mindset: treat the platform like any user-facing product with clear ownership, roadmap, and feedback loops. Shows how this approach improves developer experience and delivery speed.

An introduction to platform engineering — building shared tools and guardrails so developers can ship safely and focus on product work. Explains scope, benefits, and how to start small and evolve.