Senior (L5) · Staff (L6) · System design

ML Platform Engineer

ML platform loops blend data engineering and serving: point-in-time correctness, freshness versus cost, safe model rollout, and batch plus streaming pipelines that agree with each other.

How the loop runs

Roughly 37 minutes across three rounds.

1. 1

   High-level design

   15 min

   Batch vs streaming, storage tiers, serving path

2. 2

   Chaos injection

   12 min

   Late events, skew, hot entities, backfills

3. 3

   Cost & edge cases

   10 min

   Freshness budgets, compute cost, monitoring

Where candidates lose points

• Training and serving computing features differently
• No backfill plan for new features
• Ignoring late-arriving data