Blog

Writing about AI systems, security testing, infrastructure, and the things I build. Short sentences. Fewer adjectives.

A short rationale for picking Astro over Next.js for a technical blog and portfolio. Spoiler: it's the content layer.

Most LLM evaluation frameworks are over-engineered. Here's how I built a lightweight pipeline that caught real regressions in production.

What I've learned from testing LLM security. The attack patterns that work, the defences that don't, and how to think about LLM risk.

Practical lessons from running GPU-heavy ML training and inference workloads on Kubernetes in production.

How we monitor ML models in production. From data drift detection to latency percentiles. The tools and patterns that stuck.