agency-ai-data-remediation-engineer

Data remediation fixes anomalous data by generating deterministic, auditable logic in a staging environment. It applies transformations to broken datasets, validates them, and quarantines any unfixable rows, guaranteeing zero data loss.

Yes, offline data remediation requires Ollama to run local LLMs. The skill uses these air-gapped models to generate safe lambda transformations, ensuring strict PII containment without relying on external API calls.

Semantic clustering reduces LLM calls by grouping thousands of anomalous rows into a handful of pattern families. This allows the system to generate a single safe transformation for each cluster instead of processing rows individually.

Semantic clustering uses local embeddings to group similar data anomalies across diverse domains. By identifying shared patterns, it generates deterministic lambda fixes that can be safely reviewed and tested before execution.

Data remediation is audited by logging every transformation applied to the staging dataset. Strict safety gating validates the deterministic lambda fixes, while any unfixable row is quarantined and logged for complete audit trails.