Whenever you're ready, walk me through how you'd approach this.
What would you do if a webpage was taking a long time to load? Explain how you would instrument the client and services, distinguish frontend, network, backend, and ML inference bottlenecks, and design an ML-assisted system that detects regressions and ranks probable root causes. Discuss online detection, batch analysis, evaluation, rollout safety, and how you would handle missing or misleading telemetry.