How many GPUs do you need to serve 1,000 requests/sec, walk me through the capacity math.
The back-of-envelope chain every inference-platform round expects: traffic → tokens/sec → per-GPU throughput from bandwidth math → fleet size → dollars. With the headroom factors candidates forget.
Updated Sep 2026 · Grounded in real Forward Deployed Engineer interview loops and written to a senior-engineer editorial bar.
The back-of-envelope chain every inference-platform round expects: traffic → tokens/sec → per-GPU throughput from bandwidth math → fleet size → dollars. With the headroom factors candidates forget.
Lead with where the obvious approach breaks, because that is the judgment they are screening for — most candidates jump straight to the happy path and lose the room.
Then walk the failure back through the pipeline in order, naming the one metric the customer's exec sponsor actually cares about before you propose the fix.
