Design Saved Places - System Design | WinJob

Problem Statement: Saved Places & Address Book

Uber, Google Maps, and Apple Wallet interviews treat saved places as a read-heavy personalization layer—not a maps CDN clone. Riders pin Home, Work, gym, and airport terminals; couriers reuse drop-off bays; fleet admins maintain depot shortcuts. The hard parts are geocode normalization, deduplicating pins within ~15 meters, multi-device sync with offline edits, and never leaking another user's exact coordinates in APIs.

Phase	understanding	Section	sec-01
Mechanism	canonical place_id + user overlay

Operational metrics

list_p95 45ms
writes 800/s peak
500M rows

Failure modes to mention aloud

Duplicate pins after GPS drift—dedup by geohash7 within 15 meters.
Mobile retry without Idempotency-Key creates twin favorites—return original 200 body.
Stale etag after cross-device delete—client must accept 200 with new etag header.

Deep dive

When Uber or Google interviewers press on problem statement: saved places & address book, anchor on canonical place_id + user overlay rather than drawing unrelated microservices. Cite list p95 45ms and shard-by-user_id before naming cloud SKUs.

javaOne Dark Pro

1 public record SavedPlaceId(String value) { static SavedPlaceId parse(String raw) { return new SavedPlaceId(raw); } }

pythonOne Dark Pro

1 PlaceId = NewType("PlaceId", str)
2  
3 def merge_places(local: list[Place], remote: list[Place]) -> list[Place]:
4     """canonical place_id + user overlay merge for sec-01."""
5     return remote

typescriptOne Dark Pro

1 export type PlaceId = string & { readonly brand: unique symbol };
2 export function etagForPlaces(places: readonly SavedPlace[]): string {
3   return places.map((p) => p.version).join(":");
4 }

Why interviewers care

Saved Places interviews reward crisp scope, explicit trade-offs, and failure stories—not generic microservice diagrams.

Interview checkpoint

Name one failure story for Problem Statement: Saved Places & Address Book that proves you understand real outages, not happy-path diagrams.

Key Highlights

•canonical place_id + user overlay
•list_p95 45ms
•writes 800/s peak
•500M rows
•understanding phase checkpoint

1	PlaceId = NewType("PlaceId", str)
2
3	def merge_places(local: list[Place], remote: list[Place]) -> list[Place]:
4	"""canonical place_id + user overlay merge for sec-01."""
5	return remote

1	export type PlaceId = string & { readonly brand: unique symbol };
2	export function etagForPlaces(places: readonly SavedPlace[]): string {
3	return places.map((p) => p.version).join(":");
4	}