Free prep resources

Consolidate. Simulate. Then taper.

The final week should improve retrieval and execution, not maximize new coverage. Use one early simulation, narrow repairs, explicit risk gates, and protected recovery.

Subscribe to get the printable workbook by email. Linked examples and blank tracking sheets. Print your saved map below for your own routes and dates.

0 of 29 complete

7–5 days before · Simulate earlyRun one realistic role-specific simulation.Write the three highest-impact failures as observable behaviors.Freeze the topic list unless simulation evidence reveals a critical omission.Schedule final retries with at least one day between attempts.Confirm every expected round has a recent full attempt.If the loop includes a work sample, run that exact format rather than another generic round.Confirm six stories cover ambiguity, failure, impact, conflict, prioritization, and leadership.

4–3 days before · Repair narrowlyRetry the weakest critical answer before reviewing notes.Repair one rubric dimension at a time, not an entire domain.Practice one long round at full timing if it was weak in simulation.Rehearse story openings and ownership boundaries, not memorized prose.Prepare two thoughtful questions about the role, team, or technical environment.

2 days before · Reduce volumeRun one short mixed session; stop while retrieval is sharp.Review concise outlines, formulas, and known failure points only.Avoid a second full-loop simulation or an all-night question sprint.Confirm interview time, time zone, format, contacts, required links, and AI policy for each technical round.Open every supplied codebase, notebook, editor, or presentation link once.Test camera, microphone, screen sharing, editor, and backup connection if remote.

1 day before · Protect performanceDo at most two light confidence-building answers.Set up a quiet space, water, charger, notebook, and permitted materials.Prepare a concise role narrative: why this work, why this level, why now.If required, export the final deck, artifact links, 100-word work statement, and reference contacts.Stop preparation at a defined time and protect sleep.Treat nervousness as activation, not a signal to add content.

Interview day · Execute the processJoin or arrive early enough to resolve logistics without rushing.Clarify the objective and constraints before solving.State an outline, make decisions explicit, and invite alignment.When stuck, expose assumptions and reason forward instead of going silent.After each round, write only essential notes and reset for the next one.

Role-specific final checks

Applied Scientist

One ML design answer connects product decision, metrics, and operational reality.
Project evidence shows both scientific depth and shipping judgment.
Experiment answers include guardrails and practical ship criteria.

Machine Learning Engineer

At least one full ML implementation attempt runs correctly under time.
Design answers cover rollback, monitoring, ownership, and cost.
Debugging starts with measurements and a systematic isolation order.

Research Engineer

Implementation and systems answers quantify bottlenecks.
Research critique distinguishes claims from evidence and alternatives.
Project deep-dive connects experiments to reliable infrastructure.

Research Scientist

Derivations and mechanism answers survive changed assumptions.
Research stories separate claim, evidence, alternative, and limitation.
Work samples show discriminating experiments rather than broad exploration.

Staff / Principal

Strategy connects problem selection to opportunity cost and reversal evidence.
Architecture names decision rights, interfaces, and recovery boundaries.
Stories show durable influence and other leaders carrying the work.

Make the risk decision explicitly

Proceed

Every critical round is attempted; no critical round remains consistently Weak; remaining misses are bounded.

Proceed with known risk

One inconsistent area has a concrete repair and spaced retry before the loop. Do not pretend the risk is gone.

Extend or move when possible

More than one critical round remains Weak, ML implementation or design cannot reach a baseline in time, or story ownership collapses under follow-up.

Three things not to do

Do not replace sleep with low-quality coverage.
Do not memorize full scripts that break under follow-up.
Do not reinterpret one weak attempt as evidence that everything is weak, or one fluent repetition as proof of readiness.

If time collapses

Keep four actions: one mixed simulation, repair its two largest failures, rehearse the six-story bank, and verify logistics. Remove new topics first.

Open simulations → · Review due retries → · Practice method →