Data QA Engineer

ClearGrid is revolutionizing the debt resolution industry with cutting-edge AI and advanced analytics. We leverage real-time data, machine learning, and automation to optimize debt recovery for financial institutions. As we scale our technology platform, we are building high-performance, resilient systems to support our next phase of hyper-growth.

ClearGrid runs collections journeys across AI voice, human agents, SMS, WhatsApp and email for lenders in the UAE and KSA. Every decision — who we call, when we call them, what a promise-to-pay is worth, what we invoice a client — sits on top of data that moves from MongoDB and telephony providers into BigQuery, through dbt, and out into Tableau dashboards, Streamlit apps and client-facing reports. When that data is wrong, the consequences are not cosmetic: we call a borrower who has already paid, we mis-state a lender's recovery rate in a QBR, or we bill against the wrong call window. This role exists to stop that happening — and to catch it fast when it does.

• Test coverage in dbt across staging, intermediate and mart layers — not just not_null and unique, but business-rule tests that encode how collections actually works (DPD bucket transitions, PTP date logic, disposition hierarchies, currency and amount sanity checks).

• Source-to-warehouse reconciliation between MongoDB, telephony systems, client files and BigQuery: row counts, per-lender counts, per-lender-per-DPD-bucket counts, and orphan/drift detection.

• Freshness, volume and schema-drift monitoring on critical tables, with clear ownership of what "broken" means for each.

• Inbound client file QA — validating lender account files (fields, formats, phone numbers, names, dates, encodings) before they ever reach a journey.

• Report and dashboard QA — every Tableau dashboard and client report gets checked before it ships. Do the headline numbers reconcile to the warehouse? Do the filters, date ranges and parameters change the figures the way they should? Do totals hold when you slice by lender, by campaign, by DPD bucket? Are nulls being silently dropped from a denominator? Does the same metric agree across two dashboards that both claim to show it?