Design Address Autocomplete - System Design | WinJob

Problem Statement: Address Autocomplete at Ride-App Scale

Google, Mapbox, and HERE interviews frame autocomplete as a latency-critical read path: every keystroke triggers a ranked suggestion list biased by GPS viewport, language, and recent rides. The product goal is trustworthy pickup/drop-off selection in under 150ms p95 while controlling third-party geocoding cost via session tokens and aggressive edge caching of hot prefixes.

Phase	understanding	Section	sec-01
Mechanism	prefix index + geo bias + provider session tokens

Operational metrics

suggest_p95 120ms
14k QPS avg
8 keystrokes/session

Failure modes to mention aloud

Stale viewport bias after user pans map—refresh location_bias each suggest call.
Provider session token reuse across users—bill shock and wrong attribution.
Returning airport IATA codes without city disambiguation confuses riders.

Deep dive

javaOne Dark Pro

1 public record SuggestQuery(String q, double lat, double lng, String sessionToken) {
2   public String cacheKey() { return q.toLowerCase() + ":" + Math.round(lat * 100) + ":" + Math.round(lng * 100); }
3 }

pythonOne Dark Pro

1 SuggestQuery = TypedDict("SuggestQuery", {"q": str, "lat": float, "lng": float, "session": str})
2  
3 def cache_key(q: str, lat: float, lng: float) -> str:
4     return f"{q.lower()}:{round(lat,2)}:{round(lng,2)}"

typescriptOne Dark Pro

1 export interface SuggestQuery { q: string; lat: number; lng: number; sessionToken: string }
2 export function cacheKey({ q, lat, lng }: SuggestQuery): string {
3   return `${q.toLowerCase()}:${lat.toFixed(2)}:${lng.toFixed(2)}`;
4 }

Why interviewers care

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

Interview checkpoint

Name one failure story for Problem Statement: Address Autocomplete at Ride-App Scale that proves you understand real outages, not happy-path diagrams.

Key Highlights

•prefix index + geo bias + provider session tokens
•suggest_p95 120ms
•14k QPS avg
•8 keystrokes/session
•understanding phase checkpoint

1	public record SuggestQuery(String q, double lat, double lng, String sessionToken) {
2	public String cacheKey() { return q.toLowerCase() + ":" + Math.round(lat * 100) + ":" + Math.round(lng * 100); }
3	}

1	SuggestQuery = TypedDict("SuggestQuery", {"q": str, "lat": float, "lng": float, "session": str})
2
3	def cache_key(q: str, lat: float, lng: float) -> str:
4	return f"{q.lower()}:{round(lat,2)}:{round(lng,2)}"

1	export interface SuggestQuery { q: string; lat: number; lng: number; sessionToken: string }
2	export function cacheKey({ q, lat, lng }: SuggestQuery): string {
3	return `${q.toLowerCase()}:${lat.toFixed(2)}:${lng.toFixed(2)}`;
4	}