Notes on AI/ML engineering — LLMs, retrieval, evaluation, infrastructure, and agents.https://mosenia.com/en-ushttps://mosenia.com/posts/serving-llms-on-a-budget/https://mosenia.com/posts/serving-llms-on-a-budget/Practical patterns for running open-weight models in production when you don't have a research cluster behind you.Thu, 02 Apr 2026 00:00:00 GMTLLMsInfraarsalan@mosenia.com (Arsalan Mosenia)https://mosenia.com/posts/retrieval-is-mostly-data-work/https://mosenia.com/posts/retrieval-is-mostly-data-work/After a year of RAG systems, the pattern is clear: the interesting problems are upstream of the vector database.Tue, 24 Mar 2026 00:00:00 GMTRAGLLMsarsalan@mosenia.com (Arsalan Mosenia)https://mosenia.com/posts/building-an-eval-loop-you-trust/https://mosenia.com/posts/building-an-eval-loop-you-trust/The fastest way to ship a bad model is to grade it with a worse one. A practical guide to evaluation that survives contact with production.Wed, 11 Mar 2026 00:00:00 GMTEvalsLLMsSecurityarsalan@mosenia.com (Arsalan Mosenia)https://mosenia.com/posts/gpu-pools-without-kubernetes/https://mosenia.com/posts/gpu-pools-without-kubernetes/You probably don't need a scheduler. You need a queue, a health check, and a way to drain a box gracefully.Fri, 27 Feb 2026 00:00:00 GMTInfraLLMsarsalan@mosenia.com (Arsalan Mosenia)https://mosenia.com/posts/agents-need-boring-tools/https://mosenia.com/posts/agents-need-boring-tools/The interesting failure modes in agent systems come from tool surfaces that look fine to a human and confusing to a model.Sat, 14 Feb 2026 00:00:00 GMTAgentsSecurityEnterprise AIarsalan@mosenia.com (Arsalan Mosenia)https://mosenia.com/posts/latency-budgets-for-rag/https://mosenia.com/posts/latency-budgets-for-rag/If you haven't written the budget down, the model is deciding it for you. Here is the template I use.Fri, 30 Jan 2026 00:00:00 GMTRAGInfraarsalan@mosenia.com (Arsalan Mosenia)