Practical articles on AI, DevOps, Cloud, Linux, and infrastructure engineering.
How a packet actually gets from the internet to a pod, walked layer by layer. Plus the things that surprise people the first time they hit them.
We've shipped three end-to-end ML systems. The pieces that look obvious in slides and turn out to be the actual work.
We started routing 90% of LLM traffic through a small internal gateway. The gateway wasn't planned — it emerged from solving the same problem in 5 places. Here's the shape it took.
Prompt injection, data leakage, jailbreaks, and the boring controls that actually keep production AI features safe. The threat model that matters once you ship.
We benchmarked six embedding models on the same retrieval task. The results that surprised us, and how we'd pick today.
We cut our monthly LLM bill from $11,200 to $2,300 with seven specific changes. The ones that worked, the ones that didn't, and what we'd do first.
Fine-tuning is rarely the right answer. We've fine-tuned three times in two years; few-shot or RAG was correct for everything else. The decision criteria.
Standard APM doesn't tell you when your LLM-powered features are silently degrading. The signals we track and the dashboards that catch the regressions standard tools miss.
Multi-agent systems are mostly hype. The patterns we've seen actually deliver value, plus the ones we'd avoid until the tooling is more mature.
We have ~40 prompts in production. The patterns that improved quality, the ones that turned out to be folklore, and how we test prompts now.
How we deploy LLM-powered features. The deployment patterns are mostly normal; the validation is where the differences are.
We tried four quantization techniques on Llama-3 and Mistral models. The quality vs cost trade-offs we found, plus what works for production inference.
