Start with a closed-book baseline
Do not read the full path first.
Attempt the first scored question now. Use the result to decide which later steps need repair. Start first practice- 01
Confirm the loop
Ask whether implementation means general algorithms, ML primitives, debugging, or all three.
- 02
Establish an ML implementation baseline
Build mergeable model-evaluation logic with tests, edge cases, and explicit metric semantics.
- 03
Add bounded-memory implementation
Reason about vectorization, memory, and exact-versus-ANN trade-offs.
- 04
Debug systematically
State the diagnostic order before editing code.
- 05
Design reliable serving
Cover point-in-time data, thresholds, model outcomes, alerts, ownership, and rollback.
- 06
Make cost explicit
Quantify workload and quality, cost, and latency frontiers.
- 07
Debug offline and online disagreement
Separate leakage, experiment integrity, serving skew, proxy mismatch, and feedback effects.
- 08
Prepare failure and recovery stories
Use failure, incident recovery, trade-off, and cross-team stories.
- 09
Simulate the loop
Run ML implementation, design, production, breadth, and project rounds.
After the core path
Add only the format and subject that recruiting confirmed. Do not add every frontier lab or specialist topic.
Choose one domain supplement · Add the upper-IC level overlay · Check current lab formats · Run the role simulation