Meqyas AI

Meqyas AI is an autonomous, dialect-aware customer retention and sentiment intelligence engine for GCC e-commerce. Powered by Meta Llama 3 open weights, it decodes nuanced Gulf and Saudi Arabic dialects, detects micro-frustrations before churn happens, and automatically synthesizes personalized, budget-safe retention remedies in real time. In GCC e-commerce, high Customer Acquisition Cost (CAC) clashes with fragile retention: a single spilled container or delayed delivery leads to immediate churn. Traditional NLP tools suffer from dialect and sarcasm blindness (failing on Saudi vernacular and subtle sarcastic complaints). Meqyas AI resolves this with an autonomous multi-agent pipeline: 1. Real-Time Ingestion: Captures customer feedback across store webhooks (Salla, Zid, Shopify) and CX surveys. 2. Dialect & Sarcasm Analysis (CXRetentionAgent): Uses Meta Llama 3 to classify regional dialects (Najdi, Hijazi, Gulf), detect sarcasm, and isolate root causes (packaging spill, courier delay). 3. Calibrated Churn Remediation: Computes churn probability and Customer Lifetime Value (CLV) to autonomously formulate remedies (wallet credit, replacement, VIP escalation). 4. Deterministic Safeguards: Mathematical clamping in Python prevents compensation from exceeding order margins. Live Production URL: https://meqyas.wardat.xyz/ Full Test Suite: 83 unit tests passing in <3s.