GO TO A SECTION

↑ ↓ to navigate · Enter to openExplore all sections →
FDEInterviews logo
Challenge · Evaluation and release · Premium

Tune the production judge

Calder's live grounding judge pages so often that on-call muted it. Backtest the monitor on a recorded week, tune sample share, threshold and window to meet three constraints, then show the setting holds on weeks you did not tune on.

  • advanced
  • about 35 min
  • 5 objectives
  • par 11 steps
  • Calder Equipment
Mission workspace

This mission is part of premium, with every challenge and incident in the lab. The brief below is the whole problem; the box, the hints and the debrief are what you unlock.

Get full Premium access · ₹2,000 / $25

Every answer, concept and course, every paper breakdown, all published video lessons, hands-on FDE Lab missions, Premium PDF guides and companion files, the full practice-test bank and work-sample downloads. Referral Premium excludes guide PDFs and their companion files.

6 months · One payment · No auto-renewal

Study alongside free video lessons.

Brief
Customer
Calder Equipment (synthetic customer from the Production AI Agents course; every count and rate on this box comes from a seeded simulation made up for the exercise)
Situation
A validated grounding judge (TPR 0.88, TNR 0.97) scores a sample of live support conversations and pages when the FAIL rate over a window crosses a threshold. It pages on ordinary hours, the channel is muted, and a prompt change that raised true failures from 5% to 9% was found a day late by the human review queue.
Your role
You own the monitor. You can edit monitor.yaml and backtest it against the recorded week, or replay it across fifty other recorded weeks.
Goal
Meet all three constraints in CONSTRAINTS.md on the recorded week, explain why the starting config pages, and show your setting holds on at least 35 of 50 other weeks.
Objectives
  1. 1.Predict the false alerts the starting config raises in a normal week (you commit to an answer)
  2. 2.Name why the starting config pages on normal hours (you commit to an answer)
  3. 3.Meet all three constraints on the recorded week
  4. 4.Name the lever that buys precision with detection time instead of judge calls (you commit to an answer)
  5. 5.Show your setting holds on at least 35 of 50 other weeks