MLOps interview questions and answers

Latency is how long one request takes from arrival to response, often summarized as p50 or p99.

Throughput is how many predictions you complete per second, often QPS.

p99 is the latency below which 99 percent of requests finish

QPS is queries per second, a throughput measure of request rate.

Model serving is the production system that loads a registered model and answers prediction requests.

An endpoint decouples the model from each product service and lets you scale, version, and auth in one place.

Experiment tracking records hyperparameters, metrics, code version, and artifacts for each training attempt.

An artifact store is object storage for models, plots, datasets, and pipeline outputs, addressed by run ID or hash.

Rollback switches serving back to a previously known-good registry version, often by moving the production alias.

Training-serving skew is when the model is trained on features computed one way and served with features computed another way.

I would package the scaler, encoder, and feature builder with the model, or better, call the same library from both jobs.

It is acceptable if the feature logic is exported to a portable form, such as a feature store, ONNX, or generated code, not rewritten by hand.

I look at model size, p99 budget, QPS, and cost.

Each version should store Git SHA, DVC or table snapshot IDs, training run ID, metrics, and the serving image digest.

A signature declares expected input names, types, and shapes so serving can reject bad payloads early.

Use namespaced registered models, IAM on who can transition stages, and required reviews on production.

The offline store holds historical feature tables for training, usually warehouse or lake files.

When you build a training row, you must use feature values that were known at that event's timestamp, not future updates.

Feast lets you declare feature views, materialize to an online store, and retrieve vectors for training or serving.

A tabular feature store serves named numeric or categorical columns keyed by entity.