AI Anomaly Detection

6 Interactive Sessions

Short, interactive sessions: watch it work, try it, then build it yourself. You can go deeper at any time in the written chapters.

1. 1

   Collect clean data before you detect anything

   A detector can only find problems in data that arrived clean. Four sources, with different shapes and different clocks, have to become one tidy set of hourly readings first.

2. 2

   Learn what normal looks like, hour by hour

   A reading is only unusual compared with what is normal for that metric at that hour. The system learns the usual value and the usual wobble, then measures how far each reading sits from them.

3. 3

   Use several detectors, and check each one

   Two machine-learning detectors, an isolation forest and an autoencoder, join the z-score. Each one makes its own kind of mistake, so the system only trusts a flag when the weighted vote says so, and you check every detector against known problems.

4. 4

   One incident, one alert

   Turn a pile of anomalies into a few clear alerts that reach the right person, so a real incident is hard to miss and the on-call team keeps trusting the alerts.

5. 5

   Build the screen the on-call engineer uses

   Make the system visible and usable: an engineer opens the dashboard during an incident, sees what is wrong in seconds, and acknowledges, resolves or silences the alert.

6. 6

   From an alert to a likely cause

   Turn "search latency is three times normal" into "deploy-048, shipped just before the slow hour, is the likely cause: consider rolling it back", by joining the anomaly with recent deploys and past incidents. Then measure the monitor itself so it gets better over time.

How this certificate is scored

Your certificate shows how many of your written answers met the rubric below, checked by an AI grader. Anyone who checks your certificate sees it.

• You answer 6 short questions, one for each module, in your own words (at least 40 words each).
• An AI grader checks each answer against what its module asks for: it must answer the question, use the course's practice case specifically, and be correct.
• To pass, at least 5 of the 6 answers must meet that bar. The certificate says the answers were graded by AI.

The rubric

1. Module 1, Data Ingestion: Multi-source metric pipeline with windowing.
2. Module 2, Statistical Baselines: Moving averages, z-scores, seasonal decomposition.
3. Module 3, ML Detection: Isolation forest + autoencoder ensemble.
4. Module 4, Alert Engine: Rules, dedup, escalation, notification.
5. Module 5, Monitoring App: Next.js dashboard with anomaly timeline.
6. Module 6, Production Pipeline: Root cause analysis + SLA tracking.

What you walk away with