AI Intelligence & Anomaly Inspector

Real-time Scikit-Learn IsolationForest outlier detection & XGBoost SHAP feature attributions

Eviction Policy:AUTO-ENABLED

Cluster Node Topology — Click Node to Inspect Specific Telemetry

CPU ×4

node-ingress-01

API Gateway

🟢

CPU28%

MEM41%

44°C120W

GPU 80GB

node-worker-01

NVIDIA H100 Inference

🟢

CPU42%

MEM52%

55°C45W

CPU ×8

node-worker-02

Celery Task Runner

🟢

CPU38%

MEM48%

50°C195W

CPU ×4 NVMe

node-db-01

PostgreSQL Primary

🟢

CPU60%

MEM68%

48°C190W

Isolation Forest Anomaly Plot — node-worker-01

Triton GPU Hardware Thermal vs. Memory Allocation Plot

Normal Telemetry Cluster

Outlier Anomaly Flag

Scikit-Learn IsolationForest Model

SHAP Feature Attributions

Model decision weights for XGBoost anomaly scoring

Primary Anomaly Drivers:

1. Target Latency > 500ms (Weight: 92%)
2. Historical Error Rate = 0.85 (Weight: 85%)

💡 How to Explain This Page to Hackathon Judges & Reviewers:

1. Cluster Telemetry (Top Cards)

Shows the real-time resource utilization (CPU %, Memory %, Temp °C, Power W) across your 4 core system nodes.

2. Isolation Forest (Scatter Plot)

Detects abnormal server behavior by isolating outliers in blue (normal operational cluster) vs red (anomalous spike/failure).

3. SHAP Explainability (Right Chart)

Explains why the AI flagged an anomaly — identifying target latency and error rate as the top mathematical drivers.