Design Evernote - System Design | WinJob

Problem Statement: Evernote as a Capture-First Note Platform

Evernote is a capture-first note platform: users jot rich text, photos, PDFs, and voice memos into notebooks, then expect those artifacts to appear on every device with sub-second search. Interviewers probe whether you separate note metadata (title, notebook, tags, ACL) from content blobs (ENML/HTML bodies, attachment bytes) and whether you can explain offline-first sync with explicit conflict branches—not a naive "last writer wins" on whole files.

Production Evernote historically stressed multi-device sync, OCR inside images, and notebook hierarchies. Your design must show how a phone queues mutations in a local journal, uploads attachment chunks idempotently, and reconciles server revision numbers when the laptop reconnects after a flight.

Core actors

Consumer captures notes on mobile/desktop; expects background sync and full-text search including handwriting in photos.
Power user organizes with stacks, notebooks, and tag taxonomies; shares business notebooks with colleagues.
Enterprise admin may require encryption at rest, legal hold, and SCIM provisioning.

Why Apple/Microsoft ask this variant

They want proof you understand sync invariants: device cursor, server revision, content-addressed attachments, and searchable denormalized indexes—without conflating note bodies with attachment storage.

Design lens (problem)

Treat metadata revision order as CP per note while search/OCR stay AP with bounded lag. Quantify sync and index fanout before picking managed services. When challenged on failures, degrade search freshness and OCR backlog before blocking save acks. Tenant isolation: every storage and index query filters by user_id / tenant_id.

javaOne Dark Pro

1 public final class EvernoteproblemGuard {
2   public boolean mayApply(long baseRev, long newRev, String userId, String noteOwner) {
3     if (baseRev > newRev) return false;
4     return userId.equals(noteOwner) || aclAllows(userId, "WRITE");
5   }
6 }

pythonOne Dark Pro

1 def shard_for_user(user_id: str, shard_count: int = 256) -> int:
2     import hashlib
3     digest = hashlib.sha256(user_id.encode()).hexdigest()
4     return int(digest, 16) % shard_count

typescriptOne Dark Pro

1 export interface NoteMutation {
2   noteId: string;
3   baseRev: number;
4   clientMutationId: string;
5   deviceId: string;
6   bodyDelta?: unknown;
7 }

Why interviewers care

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

Interview checkpoint

Name one failure story for Problem Statement: Evernote as a Capture-First Note Platform that proves you understand real outages, not happy-path diagrams.

Key Highlights

•sec-001: problem — metadata rev log is CP; search/OCR are AP with SLOs.
•Evernote problem keeps attachment bytes out of the SQL hot path.
•Under load, degrade index/OCR before delaying Problem Statement: Evernote as a Capture-First Note Platform save ack.

1	public final class EvernoteproblemGuard {
2	public boolean mayApply(long baseRev, long newRev, String userId, String noteOwner) {
3	if (baseRev > newRev) return false;
4	return userId.equals(noteOwner) \|\| aclAllows(userId, "WRITE");
5	}
6	}

1	def shard_for_user(user_id: str, shard_count: int = 256) -> int:
2	import hashlib
3	digest = hashlib.sha256(user_id.encode()).hexdigest()
4	return int(digest, 16) % shard_count

1	export interface NoteMutation {
2	noteId: string;
3	baseRev: number;
4	clientMutationId: string;
5	deviceId: string;
6	bodyDelta?: unknown;
7	}