HELIXAI — AI Data Systems Developer, Capstone Project (SAIT) (AI-training and labeling pipeline engineering)
Built a human-in-the-loop pipeline to transform large volumes of unstructured supply-chain text into consistently labeled, structured records for downstream risk forecasting. Designed and applied a labeling taxonomy including risk category, severity, urgency, entity/location references, source reliability, and required action while enforcing guideline adherence. Authored AI intake instructions and QA criteria, audited model outputs for accuracy and hallucination risk, and refined labeling guidance from recurring edge cases. • Ingested raw unstructured supply-chain data into Airtable using Zapier and Make • Performed classification across thousands of records with edge-case handling • Evaluated outputs for policy fit, field completeness, clarity, and hallucination risk • Implemented instant SMS alerting when labeled thresholds were crossed