Design a high-throughput pipeline that receives raw GPS pings from a huge fleet of mobile devices, validates and enriches them, and makes a clean stream available to several downstream consumers (live tracking, geofencing, ETA, analytics).

Operating context. Device SDKs send batched pings over an authenticated edge endpoint. Write volume is enormous and bursty, devices report on cadences from one to thirty seconds, and pings arrive late, duplicated, and occasionally spoofed. Downstream consumers each read the cleaned stream independently and at their own pace. Recent per-device tracks must be queryable quickly; older data is archived cheaply.

Out of scope. The consumer services themselves (geofencing, ETA, heatmaps are separate systems), the mobile SDK internals, fine-tuning of map-matching accuracy, and long-term data-warehouse schema design. Assume the enrichment steps call existing map-match and reverse-geocode dependencies.

What to produce. A high-level architecture covering: the edge ingest API with batching, auth, and ack semantics; a durable buffer/log and its partitioning key; the validation-and-enrichment stage (dedup, outlier/spoof filtering, map-match, cell tagging); hot versus cold storage tiers for tracks; fan-out to multiple independent consumers with their own offsets; and delivery semantics and backpressure under a consumer or broker failure. Sketch the components and the journey of one ping from device to consumer; we will probe specifics in checkpoints.