Everything Is Green and the CEO Is Furious: Incident Response When the Model Is the BugJul 23, 2026·16 min read·23
Cost Guardrails for LLM SystemsMost systems get slow under overload. LLM systems just send you the bill — unless you build the structural defences first.Jul 28, 2026·16 min read·2
The Behaviour Was the Bug: Observability for Systems Where Stack Traces Don't HelpJul 16, 2026·15 min read·42
Your Eval Lied: The Architecture of a Real LLM Evaluation ProgramThe new model swept the golden set, the regression suite, and the LLM as judge evals. Then CSAT dropped. The eval did not lie. The eval measured exactly what it was built to measure. That was the problem.Jul 9, 2026·15 min read·4
The Agent Reasons, the Engine Remembers: Designing an Agentic Workflow EngineEvery "rogue agent" story is really an engine story in disguise: an agent that exceeded its budget operated in a system without one, and an agent that double executed a tool did so because the engine let it.Jul 7, 2026·21 min read·30
Nothing Scales the Way You Think: Designing an LLM Inference PlatformWhen the interviewer says "design an ML serving platform at scale," your infra instincts will betray you on every dimension: the bottleneck is GPU memory, output length determines latency, and the batching strategy you learned for image classifiers is worse than useless here.Jul 2, 2026·19 min read·79
The Intern with a Refund Button: Designing a Customer Support Agent at ScaleA support agent is the only employee you will ever deploy that can issue ten thousand refunds a second, has read every page of your help center, and will cheerfully invent a return policy if the retrieved documents come up short. Scale that carefully.Jun 30, 2026·14 min read·12
The Retriever Is Your Security Perimeter: Designing a Multi Tenant RAG Knowledge BaseThe moment an LLM starts narrating retrieval results, your retriever becomes your access control layer. Filter permissions anywhere above that and you have built a very sophisticated data leak with excellent citations.Jun 25, 2026·18 min read·18
Faster Than the Fingers: Designing a Real Time Coding Assistant*When the model sits in the loop of a typing user, the job stops being "make the model smarter" and becomes "make a model that is slower than the user feels faster than the user." Good luck.*Jun 23, 2026·18 min read·14
