Understanding how Xemlok prepares information for analysis is essential for understanding the reliability of its results.
The AI model does not work with raw screenshots or unstructured values — it receives a clean, normalized dataset constructed in real time.
Overview of the Pipeline
The pipeline consists of four major phases:
1
Data Extraction
Collect raw inputs from sources such as screenshots, logs, and structured exports. Extraction identifies key fields and captures raw values for downstream processing.
2
Data Normalization
Clean and standardize extracted values: normalize units, parse dates, standardize text fields, and validate numerical ranges so the dataset is consistent and machine-readable.
3
Context Building
Enrich normalized data with contextual metadata (e.g., user/session info, timestamps, source identifiers) and assemble related records to form a coherent view for the model.
4
AI Request Execution
Prepare the final payload and invoke the AI model with the normalized dataset and context. Handle responses, post-process model outputs, and integrate results back into the application.
