#303 Forward deployed engineer interviews 6.4 min
The FDE Decomposition Round: Vague Ask to a Plan
This plan sounded fine. A 53-line checker found eight holes in it.
Read the eight problems
Copy the code folder first. Read FIRST_PLAN in plans.py and list the problems you can see yourself: owners, metrics, data sources and what each task waits for. Then run python run.py and compare your list with the 8 problems it prints. Which ones did you miss?
A hint
Check the SOURCES table for which data exists and which access is pending.
Tighten the budget
In plans.py change BUDGET_DAYS from 30 to 20, then run python run.py. Predict what the second plan reports. You should see 1 problem, that it takes 25 days when the budget is 20. Change BUDGET_DAYS back to 30 afterwards.
A hint
The second plan is one chain, so its days add up to 25.
Remove the baseline
In the second plan change the metric of T1 from baseline recorded to time recorded, then run python run.py. Predict the result. You should see 1 problem, that nothing measures how things are today. Why is that check about the word baseline and not about the number of metrics? Change the metric back to baseline recorded afterwards.
A hint
The check looks for the word baseline in any metric of the plan.
Use data you may not touch yet
In the second plan change the source of T3 from chat logs to crm, then run python run.py. You should see 1 problem about access to crm being pending. Change it back, then change the days of T2 from 4 to 10 and predict the new problem. You should see that the plan takes 31 days against a budget of 30. Change the days of T2 back to 4 afterwards.
A hint
The crm entry in SOURCES has pending access, and the days of one chain add up.
Run two tasks in parallel
In the second plan make T2 independent by changing its needs from T1 to an empty list. Predict the finish time using the critical path, then run python run.py. You should see 22 days. Explain why it is less than 25.
A hint
T1 and T2 no longer wait for each other, so the longest chain is T2, T3, T4.
These tasks are for you to try on your own machine. Nothing is checked or saved.
What this does not show
Made-up customer and plans; the first plan was written with its holes on purpose. The checks cannot tell whether a metric or an owner is a good one. A real round is a conversation.
Up next in Forward deployed engineer interviews
More in Forward deployed engineer interviews
Related videos
- #302 Forward Deployed Engineer Interview: What Is Tested Forward deployed engineer interviews
- #305 Retry a Flaky API Without Duplicates: Idempotency Keys Forward deployed engineer interviews
- #306 Webhooks Done Right: Signatures, Retries and Replays Forward deployed engineer interviews
- #308 Two Systems Edit One Record: Lost Updates Explained Forward deployed engineer interviews
