Detect Change Points in a Trading Platform's Latency Telemetry
Overview
What this challenge is about.
You receive 90 days of per-millisecond latency telemetry across 12 services, plus an incident log of 14 known regressions and 22 known false-alarm-class events. Implement and tune a change-point detector (e.g., CUSUM, PELT, Bayesian online change-point detection — BOCPD). Target: detect all 14 regressions with at most 6 false alarms per service per month. Deliver the detector, calibration notebook, and a 2-page integration spec for the on-call paging system.
The Brief
What you'll do, and what you'll demonstrate.
Build a change-point detector that catches latency regressions on a calibrated false-alarm budget across 12 services.
Earning criteria — what you'll demonstrate
- Apply CUSUM / PELT / BOCPD to real telemetry
- Calibrate detectors to a stated false-alarm budget
- Validate against a labeled incident history
- Document an integration that engineering can ship
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
MLOps Engineer
Calibrated change-point detection feeding the on-call paging system is the bread and butter of MLOps engineers at financial-infrastructure companies.
This challenge sharpens
- change-point-detection
- telemetry-analysis
- calibration
Data Engineer
Building a detector that hooks into existing telemetry plumbing is a data-engineer-adjacent skill that bridges into infra observability.
This challenge sharpens
- telemetry-analysis
- time-series-analysis
- python
Data Scientist
Calibrating detection against a labeled history is the bread and butter of senior data-science work in any anomaly-detection setting.
This challenge sharpens
- change-point-detection
- anomaly-detection
- calibration