DATA / ANOMALY DETECTION

ValinX.

Find the signal in a noisy system. A traceable system-metrics pipeline that turns raw samples into anomaly scores, then retrieves the evidence behind them.

MY CONTRIBUTION

Connected metric ingestion, cleaning, Isolation Forest scoring, SQLite storage, and tool-based retrieval. Sample IDs carry the evidence through the full pipeline.

BUILT WITH

Python / scikit-learn / SQLite / LangChain / Ollama

Take it for a spin.

Select an anomaly, inspect the surrounding metrics, and follow the sample back to its source.

Opening the exhibit…

Loading a small, locally verified dataset.

FROM THE ORIGINAL SYSTEM

Behind the exhibit.

The copied metrics were cleaned, scored with Isolation Forest, stored in a fresh SQLite database, and retrieved by sample ID. The export records real scores and preserves their link to the source measurements.

Metrics and scores are exported from the local pipeline. Website explanations are prepared from retrieved records, not live LLM responses.

Read the actual execution transcript ↗
ORIGINAL RUN / 1 OF 3
01 / Retrieve raw measurements
01 / Retrieve raw measurements

CPU and memory measurements from the copied dataset, centered on the highest-ranked anomaly sample.

Inspect the exported data ↗
UP NEXT / 01Secure Agent ↗