Platform-side ML: reproducibility, feature and training infrastructure, registries and CI/CD for models, serving, drift, and the cost of a GPU fleet.
Say yes if it is true, give a stage and a rough timeline without naming companies, and use it to set a decision date rather than to imply a bidding war. Inventing a process you do not have is the one answer that can end badly.
Scale on queue depth or in-flight concurrency rather than request rate, because a GPU replica that takes minutes to load weights cannot be added after the queue has already formed. Then decide deliberately how much warm capacity to keep, since that is what you are paying to avoid cold starts.
Backpropagation applies the chain rule backwards through a network, reusing each layer's cached activations to get every parameter's gradient in one pass. Gradients vanish when many small factors multiply, and it shows as a stalled loss with early-layer weights barely moving while the last layers still train.
Freshness decides it. If a prediction is still valid hours later, precompute it in a batch job and serve a lookup; if it depends on what the user did seconds ago, score it online; if it depends on an evolving event sequence, score it from a stream. Most mature systems end up hybrid.
Work backwards from the review capacity. Fix the alert budget the team can clear, evaluate precision and recall at exactly that budget on data at the real prevalence, define the label and its arrival delay before modelling, and never report accuracy, which a do-nothing model scores 99.95% on.
Because you cannot score a million-item catalogue per request inside a few tens of milliseconds. Retrieval reduces millions to hundreds with a cheap model judged on recall, then a far more expensive ranker orders those hundreds and is judged on the ordering it produces at the top of the list.