An AI pipeline that qualifies and tiers logistics carriers from raw WhatsApp submissions.
Built from the manual onboarding framework I designed at Trella, deployed here as a working system.
Built from the manual carrier onboarding framework I designed and ran at Trella, a YC-backed logistics marketplace in Pakistan. The manual process had two problems: ops reps applying qualification criteria inconsistently, and no systematic way to tier carriers by performance. This system automates the standard path - taking a raw WhatsApp submission from a ground team and producing a compliance decision, performance score, and improvement plan with no manual effort.
Run Pipeline
Select any carrier from the dropdown and click Run Pipeline. You'll see the submission blob, then each stage of the pipeline: extraction, validation, qualification gate, and the agent's performance assessment. The preloaded carriers are pre-computed - no API key required (demo only).
Carrier Database
The database tab shows all 40 synthetic carrier profiles with their tiers and scores. Expand any row to see the underlying Scorecard 2 metrics. Carriers marked Rejected have no score - they were stopped at the compliance gate before the agent ran.
Try the hard cases
C016 and C017 are Urdu translation submissions. C018 is almost entirely abbreviations. C013 and C014 are borderline scores - one point apart on the tier threshold. C015 has all metrics poor and generates a multi-flag improvement plan.
Built by Mahlab Maniar · Stack: Next.js, TypeScript, Claude Sonnet 4.6, Vercel · GitHub