A Small Pipeline, With Explicit Boundaries: Architecture & ImplementationAI Waiter started with a focused goal: generate synthetic restaurant customer–waiter interactions that could be inspected, labeled, and evaluated—not just plausible-looking text. I wanted to experimen2d ago · 6 min readidirocco.hashnode.dev
A Passing Eval Doesn't Mean Your AI Feature Is ReliableWhen building an AI feature, it is tempting to treat a good evaluation score as the finish line. The pipeline runs, the examples look reasonable, the judge scores them highly. Time to ship? Not so fasOct 19 · 8 min readidirocco.hashnode.dev
My LLM Judge Went From 70% to 96% Agreement With Humans. Here’s What I Learned.An engineering diary about evaluating an AI waiter, debugging an LLM judge, and discovering that a better rubric can matter more than a cleverer model. When I started building the evaluation pipeline Oct 15 · 14 min readidirocco.hashnode.dev
