Interview System Designer

Structured scoring rubric — 5 dimensions, each 1–4

Example prompt: "Design a URL shortener for 100M URLs, 1B redirects/month."

1. Requirements clarification (15%)

4: Asks about scale, read/write ratio, latency SLA, geography, analytics needs
3: Asks about scale and core functional requirements
2: Minimal questions, starts designing immediately
1: No clarification at all

2. High-level design (25%)

4: Correct components, explains why each is needed, identifies tradeoffs
3: Correct components with adequate explanation
2: Missing a key component (no cache, no CDN for redirects)
1: Incorrect or fundamentally incomplete

3. Deep-dive on one component (30%) For URL shortener: ask them to deep-dive on hash collision handling.

4: Considers MD5 vs base62, collision probability, distributed ID generation (Snowflake)
3: Correct approach with some gap in collision handling
2: Partial solution, misses distributed coordination problem
1: Cannot reason about the component in depth

4. Scale and bottlenecks (20%)

4: Proactively identifies read-heavy (100:1 ratio) → Redis cache + CDN + read replicas
3: Identifies bottleneck when prompted
2: Identifies it but proposes wrong solution (vertical scaling only)
1: Cannot identify the bottleneck

5. Communication (10%)

4: Thinks out loud, structured narrative, changes direction gracefully when challenged
3: Adequately communicates reasoning
2: Hard to follow, jumps around
1: Cannot explain their own design

Pass threshold: ≥ 2.8 weighted average. Automatic fail if dimension 2 or 3 scores 1.

Engineering Competency Framework — 5 universal dimensions across L1–L5

Dimensions (apply to every track):

1. Technical depth
2. Technical breadth
3. Problem scope and impact
4. Communication and collaboration
5. Ownership and reliability

Level definitions:

L1 (Junior)
  Scope: Single well-defined task under close supervision
  Depth: Implements features in their primary language
  Ownership: Completes assigned work; asks for help appropriately

L2 (Mid)
  Scope: Single feature end-to-end, independently
  Depth: Designs and implements features with good abstractions
  Ownership: Drives features to completion; manages own timeline

L3 (Senior)
  Scope: Cross-service systems; mentors L1–L2
  Depth: Designs multi-service systems; identifies architectural risks
  Ownership: Finds and resolves cross-cutting concerns proactively

L4 (Staff)
  Scope: Domain-wide technical direction
  Depth: Defines architectural standards for a domain
  Ownership: Raises quality bar across teams; writes RFCs

L5 (Principal)
  Scope: Company-level technical strategy
  Depth: Industry-recognized expertise; shapes multi-year roadmap
  Ownership: Accountable for engineering org health

Backend track overlay at L3 (example):

• Designs schemas for multi-service data consistency
• Owns query performance on critical paths (P99 < 200ms)
• Writes runbooks and operational documentation for their services
• Reviews architecture proposals from L1–L2 engineers

Calibration process: managers place direct reports on the matrix independently before comparing. Disagreements of >1 level require a 30-min calibration discussion before communicating to the engineer.

Want me to generate the full matrix as a Notion table with per-track overlays for all four tracks?